SlideShare a Scribd company logo
Credibility: Evaluating what’s Been Learned
Training and Testing We measure the success of a classification procedure by using error rates (or equivalent success rates) Measuring success rate using training set is highly optimistic The error rate on training set is called resubstitution error We have a separate test set for calculating success error Test set should be independent of the training set Also some time to improve our classification technique we use a validation set When we hold out some part of training set for testing (which is now not used for training), this process is called holdout procedure
Predicting performance Expected success rate = 100 – error rate (If error rate is also in percentage) We want the true success rate Calculation of true success rate Suppose we have expected success rate(f) = s/n, where s is the number of success out of a total n instances For large value of n, f follows a normal distribution Now we will predict the true success rate (p) based on the  confidence percentage we want  For example say our f = 75%, then p will lie in [73.2%,76.7%] with 80% confidence
Predicting performance Now using properties of statistics we know that the mean of f is p and the variance is p(1-p)/n To use normal distribution we will have to make the mean of f = 0 and standard deviation = 1  So suppose our confidence = c% and we want to calculate p We will use the two tailed property of normal distribution And also that the are covered by normal distribution is taken as 100% so the are we will leave is 100 - c
Predicting performance Finally after all the manipulations we have ,true success rate as: Here,                         p -> true success rate                         f - > expected success rate                         N -> Number of instances                          Z -> Factor derived from a normal distribution table using the  100-c measure
Cross validation We use cross validation when amount of data is small and we need to have independent training and test set from it It is important that each class is represented in its actual proportions in the training and test set: Stratification  An important cross validation technique is stratified 10 fold cross validation, where the instance set is divided into 10 folds We have 10 iterations with taking a different single fold for testing and the rest 9 folds for training, averaging the error of the 10 iterations Problem: Computationally intensive
Other estimates Leave-one-out:Steps One instance is left for testing and the rest are used for training This is iterated for all the instances and the errors are averaged Leave-one-out:Advantage We use larger training sets Leave-one-out:Disadvantage Computationally intensive Cannot be stratified
Other estimates 0.632 Bootstrap Dataset of n samples is sampled n times, with replacements, to give another dataset with n instances There will be some repeated instances in the second set Here error is defined as: e = 0.632x(error in test instances) + 0.368x(error in training instances)
Comparing data mining methods Till now we were dealing with performance prediction Now we will look at methods to compare algorithms, to see which one did better We cant directly use Error rate to predict which algorithm is better as the error rate might have been calculated on different data sets So to compare algorithms we need some statistical tests We use Student’s  t- test to do this. This test help us to figure out if the mean error of two algorithm are different or not for a given confidence level
Comparing data mining methods We will use paired t-test which is a slight modification of student’s t-test Paired t-test Suppose we have unlimited data, do the following: Find k data sets from the unlimited data we have Use cross validation with each technique to get the respective outcomes: x1, x2, x3,….,xk and y1,y2,y3,……,yk mx = mean of x values and similarly my di = xi – yi Using t-statistic:
Comparing data mining methods Based on the value of k we get a degree of freedom, which enables us to figure out a z for a particular confidence value If  t <= (-z)   or  t >= (z) then, the two means differ significantly  In case t = 0 then they don’t differ, we call this null hypothesis
Predicting Probabilities Till now we were considering a scheme which when applied, results in either a correct or an incorrect prediction. This is called 0 – loss function Now we will deal with the success incase of algorithms that outputs probability distribution for e.g. Naïve Bayes
Predicting Probabilities Quadratic loss function: For a single instance there are k out comes or classes Probability vector: p1,p2,….,pk The actual out come vector is: a1,a2,a3,…..ak (where the actual outcome will be 1, rest all 0) We have to minimize the quadratic loss function given by: The minimum will be achieved when the probability vector is the true probability vector
Predicting Probabilities Informational loss function: Given by: –log(pi) Minimum is again reached at true probabilities Differences between Quadratic loss and Informational loss While quadratic loss takes all probabilities under consideration, Informational loss is based only on the class probability  While quadratic loss is bounded as its maximum output is 2, Informational loss is unbounded as it can output values up to infinity
Counting the cost Different outcomes might have different cost For example in loan decision, the cost of lending to a defaulter is far greater that the lost-business cost of  refusing a loan to a non defaulter Suppose we have two class prediction. Outcomes can be:
Counting the cost True positive rate: TP/(TP+FN) False positive rate: FP/(FP+TN) Overall success rate: Number of correct classification / Total Number of classification Error rate = 1 – success rate In multiclass case we have a confusion matrix like (actual and a random one):
Counting the cost These are the actual and the random outcome of a three class problem The diagonal represents the successful cases Kappa statistic = (D-observed  -  D-actual) / (D-perfect  -  D-actual) Here kappa statistic = (140 – 82)/(200-82) = 49.2% Kappa is used to measure the agreement between predicted and observed categorizations of a dataset, while correcting for agreements that occurs by chance Does not take cost into account
Classification with costs Example Cost matrices (just gives us the number of errors): Success rate is measured by average cost per prediction We try to minimize the costs Expected costs: dot products of vectors of class probabilities and appropriate column in cost matrix
Classification with costs Steps to take cost into consideration while testing: First use a learning method to get the probability vector (like Naïve Bayes)  Now multiple the probability vector to each column of a cost matrix one by one so as to get the cost for each class/column Select the class with the minimum(or maximum!!) cost
Cost sensitive learning Till now we included the cost factor during evaluation We will incorporate costs into the learning phase of a method We can change the ratio of instances in the training set so as to take care of costs For example we can do replication of a instances of particular class so that our learning method will give us a model with less errors of that class
Lift Charts In practice, costs are rarely known In marketing terminology the response rate is referred to as the lift factor We compare probable scenarios to make decisions A lift chart allows visual comparison Example: promotional mail out to 1,000,000 households Mail to all: 0.1%response (1000) Some data mining tool identifies subset of 100, 000 of which 0.4% respond (400) A lift of 4
Lift Charts Steps to calculate lift factor: We decide a sample size Now we arrange our data in decreasing order of the predicted probability of a class (the one which we will base our lift factor on: positive class) We calculate: Sample success proportion = Number of positive instances / Sample size  Lift factor = Sample success proportion / Data success proportion We calculate lift factor for different sample size to get  Lift Charts
Lift Charts A hypothetical lift chart
Lift Charts In the lift chart we will like to stay towards the upper left corner The diagonal line is the curve for random samples without using sorted data Any good selection will keep the lift curve above the diagonal
ROC Curves Stands for receiver operating characteristic Difference to lift charts: Y axis showspercentage of true positive  X axis shows percentage of false positives in samples ROC is a jagged curve It can be smoothened out by cross validation
ROC Curves A ROC curve
ROC Curves Ways to generate cost curves (Consider the previous diagram for reference) First way: Get the probability distribution over different folds of data Sort the data in decreasing order of the probability of yes class Select a point on X-axis and for that number of no, get the number of yes for each probability distribution Average the number of yes from all the folds and plot it
ROC Curves Second way: Get the probability distribution over different folds of data Sort the data in decreasing order of the probability of yes class Select a point on X-axis and for that number of no, get the number of yes for each probability distribution Plot a ROC for each fold individually  Average all the ROCs
ROC Curves ROC curves for two schemes
ROC Curves In the previous ROC curves: For a small, focused sample, use method A For a large one, use method B In between, choose between A and B with appropriate probabilities
Recall – precision curves In case of a search query: Recall = number of documents retrieved that are relevant / total number of documents that are relevant Precision = number of documents retrieved that are relevant / total number of documents that are retrieved
A summary          Different measures used to evaluate the false positive versus the false negative tradeoff
Cost curves Cost curves plot expected costs directly Example for case with uniform costs (i.e. error):
Cost curves Example with costs:
Cost curves C[+|-]  is the cost of predicting + when the instance is – C[-|+]  is the cost of predicting - when the instance is +
Minimum Description Length Principle The description length is defined as: Space required to describe a theory + space required to describe the theory’s mistakes Theory  = Classifier and mistakes = errors on the training data We try to minimize the description length MDL theory is the one that compresses the data the most. I.e to compress a data set we generate a model and then store the model and its mistakes We need to compute: Size of the model Space needed to encode the error
Minimum Description Length Principle The 2nd  one is easy. Just use informational loss function For  1st  we need a method to encode the model L[T] = “length” of the theory L[E|T] = training set encoded wrt the theory
Minimum Description Length Principle MDL and clustering Description length of theory: bits needed to encode the clusters. E.g. cluster centers Description length of data given theory: encode cluster membership and position relative to cluster. E.g. distance to cluster centers Works if coding scheme uses less code space for small numbers than for large ones
Visit more self help tutorials Pick a tutorial of your choice and browse through it at your own pace. The tutorials section is free, self-guiding and will not involve any additional support. Visit us at www.dataminingtools.net

More Related Content

What's hot

Linear Regression Ex
Linear Regression ExLinear Regression Ex
Linear Regression Ex
mailund
 
Machine learning session7(nb classifier k-nn)
Machine learning   session7(nb classifier k-nn)Machine learning   session7(nb classifier k-nn)
Machine learning session7(nb classifier k-nn)
Abhimanyu Dwivedi
 
Heart disease classification
Heart disease classificationHeart disease classification
Heart disease classification
SnehaDey21
 
Decision tree and random forest
Decision tree and random forestDecision tree and random forest
Decision tree and random forest
Lippo Group Digital
 
Classification Based Machine Learning Algorithms
Classification Based Machine Learning AlgorithmsClassification Based Machine Learning Algorithms
Classification Based Machine Learning Algorithms
Md. Main Uddin Rony
 
Data Science - Part IX - Support Vector Machine
Data Science - Part IX -  Support Vector MachineData Science - Part IX -  Support Vector Machine
Data Science - Part IX - Support Vector Machine
Derek Kane
 
Random Forest / Bootstrap Aggregation
Random Forest / Bootstrap AggregationRandom Forest / Bootstrap Aggregation
Random Forest / Bootstrap Aggregation
Rupak Roy
 
Machine Learning - Simple Linear Regression
Machine Learning - Simple Linear RegressionMachine Learning - Simple Linear Regression
Machine Learning - Simple Linear Regression
Siddharth Shrivastava
 
Machine learning session1
Machine learning   session1Machine learning   session1
Machine learning session1
Abhimanyu Dwivedi
 
WEKA: Algorithms The Basic Methods
WEKA: Algorithms The Basic MethodsWEKA: Algorithms The Basic Methods
WEKA: Algorithms The Basic Methods
DataminingTools Inc
 
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)Machine learning basics using trees algorithm (Random forest, Gradient Boosting)
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)
Parth Khare
 
Aaa ped-14-Ensemble Learning: About Ensemble Learning
Aaa ped-14-Ensemble Learning: About Ensemble LearningAaa ped-14-Ensemble Learning: About Ensemble Learning
Aaa ped-14-Ensemble Learning: About Ensemble Learning
AminaRepo
 
Data Analysis: Evaluation Metrics for Supervised Learning Models of Machine L...
Data Analysis: Evaluation Metrics for Supervised Learning Models of Machine L...Data Analysis: Evaluation Metrics for Supervised Learning Models of Machine L...
Data Analysis: Evaluation Metrics for Supervised Learning Models of Machine L...
Md. Main Uddin Rony
 
L4. Ensembles of Decision Trees
L4. Ensembles of Decision TreesL4. Ensembles of Decision Trees
L4. Ensembles of Decision Trees
Machine Learning Valencia
 
Machine Learning using Support Vector Machine
Machine Learning using Support Vector MachineMachine Learning using Support Vector Machine
Machine Learning using Support Vector Machine
Mohsin Ul Haq
 
Borderline Smote
Borderline SmoteBorderline Smote
Borderline Smote
Trector Rancor
 
An Introduction to Simulation in the Social Sciences
An Introduction to Simulation in the Social SciencesAn Introduction to Simulation in the Social Sciences
An Introduction to Simulation in the Social Sciences
fsmart01
 
Predict Backorder on a supply chain data for an Organization
Predict Backorder on a supply chain data for an OrganizationPredict Backorder on a supply chain data for an Organization
Predict Backorder on a supply chain data for an Organization
Piyush Srivastava
 
Data Science - Part III - EDA & Model Selection
Data Science - Part III - EDA & Model SelectionData Science - Part III - EDA & Model Selection
Data Science - Part III - EDA & Model Selection
Derek Kane
 
Linear Regression, Machine learning term
Linear Regression, Machine learning termLinear Regression, Machine learning term
Linear Regression, Machine learning term
S Rulez
 

What's hot (20)

Linear Regression Ex
Linear Regression ExLinear Regression Ex
Linear Regression Ex
 
Machine learning session7(nb classifier k-nn)
Machine learning   session7(nb classifier k-nn)Machine learning   session7(nb classifier k-nn)
Machine learning session7(nb classifier k-nn)
 
Heart disease classification
Heart disease classificationHeart disease classification
Heart disease classification
 
Decision tree and random forest
Decision tree and random forestDecision tree and random forest
Decision tree and random forest
 
Classification Based Machine Learning Algorithms
Classification Based Machine Learning AlgorithmsClassification Based Machine Learning Algorithms
Classification Based Machine Learning Algorithms
 
Data Science - Part IX - Support Vector Machine
Data Science - Part IX -  Support Vector MachineData Science - Part IX -  Support Vector Machine
Data Science - Part IX - Support Vector Machine
 
Random Forest / Bootstrap Aggregation
Random Forest / Bootstrap AggregationRandom Forest / Bootstrap Aggregation
Random Forest / Bootstrap Aggregation
 
Machine Learning - Simple Linear Regression
Machine Learning - Simple Linear RegressionMachine Learning - Simple Linear Regression
Machine Learning - Simple Linear Regression
 
Machine learning session1
Machine learning   session1Machine learning   session1
Machine learning session1
 
WEKA: Algorithms The Basic Methods
WEKA: Algorithms The Basic MethodsWEKA: Algorithms The Basic Methods
WEKA: Algorithms The Basic Methods
 
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)Machine learning basics using trees algorithm (Random forest, Gradient Boosting)
Machine learning basics using trees algorithm (Random forest, Gradient Boosting)
 
Aaa ped-14-Ensemble Learning: About Ensemble Learning
Aaa ped-14-Ensemble Learning: About Ensemble LearningAaa ped-14-Ensemble Learning: About Ensemble Learning
Aaa ped-14-Ensemble Learning: About Ensemble Learning
 
Data Analysis: Evaluation Metrics for Supervised Learning Models of Machine L...
Data Analysis: Evaluation Metrics for Supervised Learning Models of Machine L...Data Analysis: Evaluation Metrics for Supervised Learning Models of Machine L...
Data Analysis: Evaluation Metrics for Supervised Learning Models of Machine L...
 
L4. Ensembles of Decision Trees
L4. Ensembles of Decision TreesL4. Ensembles of Decision Trees
L4. Ensembles of Decision Trees
 
Machine Learning using Support Vector Machine
Machine Learning using Support Vector MachineMachine Learning using Support Vector Machine
Machine Learning using Support Vector Machine
 
Borderline Smote
Borderline SmoteBorderline Smote
Borderline Smote
 
An Introduction to Simulation in the Social Sciences
An Introduction to Simulation in the Social SciencesAn Introduction to Simulation in the Social Sciences
An Introduction to Simulation in the Social Sciences
 
Predict Backorder on a supply chain data for an Organization
Predict Backorder on a supply chain data for an OrganizationPredict Backorder on a supply chain data for an Organization
Predict Backorder on a supply chain data for an Organization
 
Data Science - Part III - EDA & Model Selection
Data Science - Part III - EDA & Model SelectionData Science - Part III - EDA & Model Selection
Data Science - Part III - EDA & Model Selection
 
Linear Regression, Machine learning term
Linear Regression, Machine learning termLinear Regression, Machine learning term
Linear Regression, Machine learning term
 

Viewers also liked

Ofimatica
OfimaticaOfimatica
Ofimatica
cpinzonsite
 
Pdf bimestral diapositivas
Pdf bimestral diapositivasPdf bimestral diapositivas
Pdf bimestral diapositivas
cpinzonsite
 
"ЖАЗУЫ CODE - түрме ӘЛЕМ Матрикс OUT РУХАНИ
"ЖАЗУЫ CODE - түрме ӘЛЕМ Матрикс OUT РУХАНИ "ЖАЗУЫ CODE - түрме ӘЛЕМ Матрикс OUT РУХАНИ
"ЖАЗУЫ CODE - түрме ӘЛЕМ Матрикс OUT РУХАНИ
cdoecrt
 
10 gode råd om samliv i landbruket
10 gode råd om samliv i landbruket10 gode råd om samliv i landbruket
10 gode råd om samliv i landbruket
Ellen Krageberg
 
Fylkeslagets rolle i bygdekvinnelaget
Fylkeslagets rolle i bygdekvinnelagetFylkeslagets rolle i bygdekvinnelaget
Fylkeslagets rolle i bygdekvinnelaget
Ellen Krageberg
 
"كتابة التعليمات البرمجية - سجن WORLD MATRIX OUT الروحية
"كتابة التعليمات البرمجية - سجن WORLD MATRIX OUT الروحية "كتابة التعليمات البرمجية - سجن WORLD MATRIX OUT الروحية
"كتابة التعليمات البرمجية - سجن WORLD MATRIX OUT الروحية
cdoecrt
 
WEKA:Practical Machine Learning Tools And Techniques
WEKA:Practical Machine Learning Tools And TechniquesWEKA:Practical Machine Learning Tools And Techniques
WEKA:Practical Machine Learning Tools And Techniques
weka Content
 
CV UK
CV UKCV UK
10 een marketingcampagne plannen in ms dynamics crm
10 een marketingcampagne plannen in ms dynamics crm10 een marketingcampagne plannen in ms dynamics crm
10 een marketingcampagne plannen in ms dynamics crmLaura Moens
 
What is the Artez Mobile Fundraising App?
What is the Artez Mobile Fundraising App?What is the Artez Mobile Fundraising App?
What is the Artez Mobile Fundraising App?
Artez Interactive
 
SolidWorks Model : SW2015X-A12
SolidWorks  Model : SW2015X-A12SolidWorks  Model : SW2015X-A12
SolidWorks Model : SW2015X-A12
Taehwan Kim
 
Presentación Francisco Lupiáñez, las TIC para la Salud
Presentación Francisco Lupiáñez, las TIC para la SaludPresentación Francisco Lupiáñez, las TIC para la Salud
Presentación Francisco Lupiáñez, las TIC para la Salud
UOCsevilla
 
Building Business Service Intelligence
Building Business Service IntelligenceBuilding Business Service Intelligence
Building Business Service Intelligence
Splunk
 
Design e conteúdo, o 'prato feito' das mídias sociais
Design e conteúdo, o 'prato feito' das mídias sociaisDesign e conteúdo, o 'prato feito' das mídias sociais
Design e conteúdo, o 'prato feito' das mídias sociais
Diego Motta
 
What is the Artez Mobile Fundraising App for iPhone?
What is the Artez Mobile Fundraising App for iPhone?What is the Artez Mobile Fundraising App for iPhone?
What is the Artez Mobile Fundraising App for iPhone?
Artez Interactive
 
WEKA:Output Knowledge Representation
WEKA:Output Knowledge RepresentationWEKA:Output Knowledge Representation
WEKA:Output Knowledge Representation
weka Content
 
WEKA: The Knowledge Flow Interface
WEKA: The Knowledge Flow InterfaceWEKA: The Knowledge Flow Interface
WEKA: The Knowledge Flow Interface
weka Content
 
HEZKIDETZA
HEZKIDETZAHEZKIDETZA
HEZKIDETZA
Nagore Ruedas
 
Variantes normales en el eeg
Variantes normales en el eegVariantes normales en el eeg
Variantes normales en el eeg
magamtzf
 

Viewers also liked (20)

Ofimatica
OfimaticaOfimatica
Ofimatica
 
Pdf bimestral diapositivas
Pdf bimestral diapositivasPdf bimestral diapositivas
Pdf bimestral diapositivas
 
"ЖАЗУЫ CODE - түрме ӘЛЕМ Матрикс OUT РУХАНИ
"ЖАЗУЫ CODE - түрме ӘЛЕМ Матрикс OUT РУХАНИ "ЖАЗУЫ CODE - түрме ӘЛЕМ Матрикс OUT РУХАНИ
"ЖАЗУЫ CODE - түрме ӘЛЕМ Матрикс OUT РУХАНИ
 
10 gode råd om samliv i landbruket
10 gode råd om samliv i landbruket10 gode råd om samliv i landbruket
10 gode råd om samliv i landbruket
 
Fylkeslagets rolle i bygdekvinnelaget
Fylkeslagets rolle i bygdekvinnelagetFylkeslagets rolle i bygdekvinnelaget
Fylkeslagets rolle i bygdekvinnelaget
 
Verkkokaupan8Tukitointa
Verkkokaupan8TukitointaVerkkokaupan8Tukitointa
Verkkokaupan8Tukitointa
 
"كتابة التعليمات البرمجية - سجن WORLD MATRIX OUT الروحية
"كتابة التعليمات البرمجية - سجن WORLD MATRIX OUT الروحية "كتابة التعليمات البرمجية - سجن WORLD MATRIX OUT الروحية
"كتابة التعليمات البرمجية - سجن WORLD MATRIX OUT الروحية
 
WEKA:Practical Machine Learning Tools And Techniques
WEKA:Practical Machine Learning Tools And TechniquesWEKA:Practical Machine Learning Tools And Techniques
WEKA:Practical Machine Learning Tools And Techniques
 
CV UK
CV UKCV UK
CV UK
 
10 een marketingcampagne plannen in ms dynamics crm
10 een marketingcampagne plannen in ms dynamics crm10 een marketingcampagne plannen in ms dynamics crm
10 een marketingcampagne plannen in ms dynamics crm
 
What is the Artez Mobile Fundraising App?
What is the Artez Mobile Fundraising App?What is the Artez Mobile Fundraising App?
What is the Artez Mobile Fundraising App?
 
SolidWorks Model : SW2015X-A12
SolidWorks  Model : SW2015X-A12SolidWorks  Model : SW2015X-A12
SolidWorks Model : SW2015X-A12
 
Presentación Francisco Lupiáñez, las TIC para la Salud
Presentación Francisco Lupiáñez, las TIC para la SaludPresentación Francisco Lupiáñez, las TIC para la Salud
Presentación Francisco Lupiáñez, las TIC para la Salud
 
Building Business Service Intelligence
Building Business Service IntelligenceBuilding Business Service Intelligence
Building Business Service Intelligence
 
Design e conteúdo, o 'prato feito' das mídias sociais
Design e conteúdo, o 'prato feito' das mídias sociaisDesign e conteúdo, o 'prato feito' das mídias sociais
Design e conteúdo, o 'prato feito' das mídias sociais
 
What is the Artez Mobile Fundraising App for iPhone?
What is the Artez Mobile Fundraising App for iPhone?What is the Artez Mobile Fundraising App for iPhone?
What is the Artez Mobile Fundraising App for iPhone?
 
WEKA:Output Knowledge Representation
WEKA:Output Knowledge RepresentationWEKA:Output Knowledge Representation
WEKA:Output Knowledge Representation
 
WEKA: The Knowledge Flow Interface
WEKA: The Knowledge Flow InterfaceWEKA: The Knowledge Flow Interface
WEKA: The Knowledge Flow Interface
 
HEZKIDETZA
HEZKIDETZAHEZKIDETZA
HEZKIDETZA
 
Variantes normales en el eeg
Variantes normales en el eegVariantes normales en el eeg
Variantes normales en el eeg
 

Similar to WEKA:Credibility Evaluating Whats Been Learned

Assessing Model Performance - Beginner's Guide
Assessing Model Performance - Beginner's GuideAssessing Model Performance - Beginner's Guide
Assessing Model Performance - Beginner's Guide
Megan Verbakel
 
working with python
working with pythonworking with python
working with python
bhavesh lande
 
Machine learning session4(linear regression)
Machine learning   session4(linear regression)Machine learning   session4(linear regression)
Machine learning session4(linear regression)
Abhimanyu Dwivedi
 
MLlectureMethod.ppt
MLlectureMethod.pptMLlectureMethod.ppt
MLlectureMethod.ppt
butest
 
MLlectureMethod.ppt
MLlectureMethod.pptMLlectureMethod.ppt
MLlectureMethod.ppt
butest
 
Lecture 3.1_ Logistic Regression.pptx
Lecture 3.1_ Logistic Regression.pptxLecture 3.1_ Logistic Regression.pptx
Lecture 3.1_ Logistic Regression.pptx
ajondaree
 
evaluation and credibility-Part 1
evaluation and credibility-Part 1evaluation and credibility-Part 1
Understanding Blackbox Prediction via Influence Functions
Understanding Blackbox Prediction via Influence FunctionsUnderstanding Blackbox Prediction via Influence Functions
Understanding Blackbox Prediction via Influence Functions
SEMINARGROOT
 
Machine learning (5)
Machine learning (5)Machine learning (5)
Machine learning (5)
NYversity
 
13ClassifierPerformance.pdf
13ClassifierPerformance.pdf13ClassifierPerformance.pdf
13ClassifierPerformance.pdf
ssuserdce5c21
 
PERFORMANCE_PREDICTION__PARAMETERS[1].pptx
PERFORMANCE_PREDICTION__PARAMETERS[1].pptxPERFORMANCE_PREDICTION__PARAMETERS[1].pptx
PERFORMANCE_PREDICTION__PARAMETERS[1].pptx
TAHIRZAMAN81
 
All PERFORMANCE PREDICTION PARAMETERS.pptx
All PERFORMANCE PREDICTION  PARAMETERS.pptxAll PERFORMANCE PREDICTION  PARAMETERS.pptx
All PERFORMANCE PREDICTION PARAMETERS.pptx
taherzamanrather
 
Cross Validation Cross ValidationmCross Validation.pptx
Cross Validation Cross ValidationmCross Validation.pptxCross Validation Cross ValidationmCross Validation.pptx
Cross Validation Cross ValidationmCross Validation.pptx
Nishant83346
 
Py data19 final
Py data19   finalPy data19   final
Py data19 final
Maria Navarro Jiménez
 
Supervised Learning.pdf
Supervised Learning.pdfSupervised Learning.pdf
Supervised Learning.pdf
gadissaassefa
 
Machine learning and_nlp
Machine learning and_nlpMachine learning and_nlp
Machine learning and_nlp
ankit_ppt
 
BOOTSTRAPPING TO EVALUATE RESPONSE MODELS: A SAS® MACRO
BOOTSTRAPPING TO EVALUATE RESPONSE MODELS: A SAS® MACROBOOTSTRAPPING TO EVALUATE RESPONSE MODELS: A SAS® MACRO
BOOTSTRAPPING TO EVALUATE RESPONSE MODELS: A SAS® MACRO
Anthony Kilili
 
Essentials of machine learning algorithms
Essentials of machine learning algorithmsEssentials of machine learning algorithms
Essentials of machine learning algorithms
Arunangsu Sahu
 
MACHINE LEARNING PPT K MEANS CLUSTERING.
MACHINE LEARNING PPT K MEANS CLUSTERING.MACHINE LEARNING PPT K MEANS CLUSTERING.
MACHINE LEARNING PPT K MEANS CLUSTERING.
AmnaArooj13
 
INTRODUCTION TO BOOSTING.ppt
INTRODUCTION TO BOOSTING.pptINTRODUCTION TO BOOSTING.ppt
INTRODUCTION TO BOOSTING.ppt
BharatDaiyaBharat
 

Similar to WEKA:Credibility Evaluating Whats Been Learned (20)

Assessing Model Performance - Beginner's Guide
Assessing Model Performance - Beginner's GuideAssessing Model Performance - Beginner's Guide
Assessing Model Performance - Beginner's Guide
 
working with python
working with pythonworking with python
working with python
 
Machine learning session4(linear regression)
Machine learning   session4(linear regression)Machine learning   session4(linear regression)
Machine learning session4(linear regression)
 
MLlectureMethod.ppt
MLlectureMethod.pptMLlectureMethod.ppt
MLlectureMethod.ppt
 
MLlectureMethod.ppt
MLlectureMethod.pptMLlectureMethod.ppt
MLlectureMethod.ppt
 
Lecture 3.1_ Logistic Regression.pptx
Lecture 3.1_ Logistic Regression.pptxLecture 3.1_ Logistic Regression.pptx
Lecture 3.1_ Logistic Regression.pptx
 
evaluation and credibility-Part 1
evaluation and credibility-Part 1evaluation and credibility-Part 1
evaluation and credibility-Part 1
 
Understanding Blackbox Prediction via Influence Functions
Understanding Blackbox Prediction via Influence FunctionsUnderstanding Blackbox Prediction via Influence Functions
Understanding Blackbox Prediction via Influence Functions
 
Machine learning (5)
Machine learning (5)Machine learning (5)
Machine learning (5)
 
13ClassifierPerformance.pdf
13ClassifierPerformance.pdf13ClassifierPerformance.pdf
13ClassifierPerformance.pdf
 
PERFORMANCE_PREDICTION__PARAMETERS[1].pptx
PERFORMANCE_PREDICTION__PARAMETERS[1].pptxPERFORMANCE_PREDICTION__PARAMETERS[1].pptx
PERFORMANCE_PREDICTION__PARAMETERS[1].pptx
 
All PERFORMANCE PREDICTION PARAMETERS.pptx
All PERFORMANCE PREDICTION  PARAMETERS.pptxAll PERFORMANCE PREDICTION  PARAMETERS.pptx
All PERFORMANCE PREDICTION PARAMETERS.pptx
 
Cross Validation Cross ValidationmCross Validation.pptx
Cross Validation Cross ValidationmCross Validation.pptxCross Validation Cross ValidationmCross Validation.pptx
Cross Validation Cross ValidationmCross Validation.pptx
 
Py data19 final
Py data19   finalPy data19   final
Py data19 final
 
Supervised Learning.pdf
Supervised Learning.pdfSupervised Learning.pdf
Supervised Learning.pdf
 
Machine learning and_nlp
Machine learning and_nlpMachine learning and_nlp
Machine learning and_nlp
 
BOOTSTRAPPING TO EVALUATE RESPONSE MODELS: A SAS® MACRO
BOOTSTRAPPING TO EVALUATE RESPONSE MODELS: A SAS® MACROBOOTSTRAPPING TO EVALUATE RESPONSE MODELS: A SAS® MACRO
BOOTSTRAPPING TO EVALUATE RESPONSE MODELS: A SAS® MACRO
 
Essentials of machine learning algorithms
Essentials of machine learning algorithmsEssentials of machine learning algorithms
Essentials of machine learning algorithms
 
MACHINE LEARNING PPT K MEANS CLUSTERING.
MACHINE LEARNING PPT K MEANS CLUSTERING.MACHINE LEARNING PPT K MEANS CLUSTERING.
MACHINE LEARNING PPT K MEANS CLUSTERING.
 
INTRODUCTION TO BOOSTING.ppt
INTRODUCTION TO BOOSTING.pptINTRODUCTION TO BOOSTING.ppt
INTRODUCTION TO BOOSTING.ppt
 

More from weka Content

WEKA:The Command Line Interface
WEKA:The Command Line InterfaceWEKA:The Command Line Interface
WEKA:The Command Line Interface
weka Content
 
WEKA:The Experimenter
WEKA:The ExperimenterWEKA:The Experimenter
WEKA:The Experimenter
weka Content
 
WEKA:The Explorer
WEKA:The ExplorerWEKA:The Explorer
WEKA:The Explorer
weka Content
 
WEKA:Algorithms The Basic Methods
WEKA:Algorithms The Basic MethodsWEKA:Algorithms The Basic Methods
WEKA:Algorithms The Basic Methods
weka Content
 
WEKA:Data Mining Input Concepts Instances And Attributes
WEKA:Data Mining Input Concepts Instances And AttributesWEKA:Data Mining Input Concepts Instances And Attributes
WEKA:Data Mining Input Concepts Instances And Attributes
weka Content
 
WEKA:Introduction To Weka
WEKA:Introduction To WekaWEKA:Introduction To Weka
WEKA:Introduction To Weka
weka Content
 
An Introduction To Weka
An Introduction To WekaAn Introduction To Weka
An Introduction To Weka
weka Content
 

More from weka Content (7)

WEKA:The Command Line Interface
WEKA:The Command Line InterfaceWEKA:The Command Line Interface
WEKA:The Command Line Interface
 
WEKA:The Experimenter
WEKA:The ExperimenterWEKA:The Experimenter
WEKA:The Experimenter
 
WEKA:The Explorer
WEKA:The ExplorerWEKA:The Explorer
WEKA:The Explorer
 
WEKA:Algorithms The Basic Methods
WEKA:Algorithms The Basic MethodsWEKA:Algorithms The Basic Methods
WEKA:Algorithms The Basic Methods
 
WEKA:Data Mining Input Concepts Instances And Attributes
WEKA:Data Mining Input Concepts Instances And AttributesWEKA:Data Mining Input Concepts Instances And Attributes
WEKA:Data Mining Input Concepts Instances And Attributes
 
WEKA:Introduction To Weka
WEKA:Introduction To WekaWEKA:Introduction To Weka
WEKA:Introduction To Weka
 
An Introduction To Weka
An Introduction To WekaAn Introduction To Weka
An Introduction To Weka
 

Recently uploaded

Authentically Social by Corey Perlman - EO Puerto Rico
Authentically Social by Corey Perlman - EO Puerto RicoAuthentically Social by Corey Perlman - EO Puerto Rico
Authentically Social by Corey Perlman - EO Puerto Rico
Corey Perlman, Social Media Speaker and Consultant
 
Company Valuation webinar series - Tuesday, 4 June 2024
Company Valuation webinar series - Tuesday, 4 June 2024Company Valuation webinar series - Tuesday, 4 June 2024
Company Valuation webinar series - Tuesday, 4 June 2024
FelixPerez547899
 
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
my Pandit
 
How to Implement a Strategy: Transform Your Strategy with BSC Designer's Comp...
How to Implement a Strategy: Transform Your Strategy with BSC Designer's Comp...How to Implement a Strategy: Transform Your Strategy with BSC Designer's Comp...
How to Implement a Strategy: Transform Your Strategy with BSC Designer's Comp...
Aleksey Savkin
 
Unveiling the Dynamic Personalities, Key Dates, and Horoscope Insights: Gemin...
Unveiling the Dynamic Personalities, Key Dates, and Horoscope Insights: Gemin...Unveiling the Dynamic Personalities, Key Dates, and Horoscope Insights: Gemin...
Unveiling the Dynamic Personalities, Key Dates, and Horoscope Insights: Gemin...
my Pandit
 
amptalk_RecruitingDeck_english_2024.06.05
amptalk_RecruitingDeck_english_2024.06.05amptalk_RecruitingDeck_english_2024.06.05
amptalk_RecruitingDeck_english_2024.06.05
marketing317746
 
Zodiac Signs and Food Preferences_ What Your Sign Says About Your Taste
Zodiac Signs and Food Preferences_ What Your Sign Says About Your TasteZodiac Signs and Food Preferences_ What Your Sign Says About Your Taste
Zodiac Signs and Food Preferences_ What Your Sign Says About Your Taste
my Pandit
 
-- June 2024 is National Volunteer Month --
-- June 2024 is National Volunteer Month ---- June 2024 is National Volunteer Month --
-- June 2024 is National Volunteer Month --
NZSG
 
Building Your Employer Brand with Social Media
Building Your Employer Brand with Social MediaBuilding Your Employer Brand with Social Media
Building Your Employer Brand with Social Media
LuanWise
 
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
Lacey Max
 
2024-6-01-IMPACTSilver-Corp-Presentation.pdf
2024-6-01-IMPACTSilver-Corp-Presentation.pdf2024-6-01-IMPACTSilver-Corp-Presentation.pdf
2024-6-01-IMPACTSilver-Corp-Presentation.pdf
hartfordclub1
 
Digital Transformation Frameworks: Driving Digital Excellence
Digital Transformation Frameworks: Driving Digital ExcellenceDigital Transformation Frameworks: Driving Digital Excellence
Digital Transformation Frameworks: Driving Digital Excellence
Operational Excellence Consulting
 
Best Forex Brokers Comparison in INDIA 2024
Best Forex Brokers Comparison in INDIA 2024Best Forex Brokers Comparison in INDIA 2024
Best Forex Brokers Comparison in INDIA 2024
Top Forex Brokers Review
 
How MJ Global Leads the Packaging Industry.pdf
How MJ Global Leads the Packaging Industry.pdfHow MJ Global Leads the Packaging Industry.pdf
How MJ Global Leads the Packaging Industry.pdf
MJ Global
 
Digital Marketing with a Focus on Sustainability
Digital Marketing with a Focus on SustainabilityDigital Marketing with a Focus on Sustainability
Digital Marketing with a Focus on Sustainability
sssourabhsharma
 
The APCO Geopolitical Radar - Q3 2024 The Global Operating Environment for Bu...
The APCO Geopolitical Radar - Q3 2024 The Global Operating Environment for Bu...The APCO Geopolitical Radar - Q3 2024 The Global Operating Environment for Bu...
The APCO Geopolitical Radar - Q3 2024 The Global Operating Environment for Bu...
APCO
 
Hamster Kombat' Telegram Game Surpasses 100 Million Players—Token Release Sch...
Hamster Kombat' Telegram Game Surpasses 100 Million Players—Token Release Sch...Hamster Kombat' Telegram Game Surpasses 100 Million Players—Token Release Sch...
Hamster Kombat' Telegram Game Surpasses 100 Million Players—Token Release Sch...
SOFTTECHHUB
 
Anny Serafina Love - Letter of Recommendation by Kellen Harkins, MS.
Anny Serafina Love - Letter of Recommendation by Kellen Harkins, MS.Anny Serafina Love - Letter of Recommendation by Kellen Harkins, MS.
Anny Serafina Love - Letter of Recommendation by Kellen Harkins, MS.
AnnySerafinaLove
 
Business storytelling: key ingredients to a story
Business storytelling: key ingredients to a storyBusiness storytelling: key ingredients to a story
Business storytelling: key ingredients to a story
Alexandra Fulford
 
The 10 Most Influential Leaders Guiding Corporate Evolution, 2024.pdf
The 10 Most Influential Leaders Guiding Corporate Evolution, 2024.pdfThe 10 Most Influential Leaders Guiding Corporate Evolution, 2024.pdf
The 10 Most Influential Leaders Guiding Corporate Evolution, 2024.pdf
thesiliconleaders
 

Recently uploaded (20)

Authentically Social by Corey Perlman - EO Puerto Rico
Authentically Social by Corey Perlman - EO Puerto RicoAuthentically Social by Corey Perlman - EO Puerto Rico
Authentically Social by Corey Perlman - EO Puerto Rico
 
Company Valuation webinar series - Tuesday, 4 June 2024
Company Valuation webinar series - Tuesday, 4 June 2024Company Valuation webinar series - Tuesday, 4 June 2024
Company Valuation webinar series - Tuesday, 4 June 2024
 
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
Taurus Zodiac Sign: Unveiling the Traits, Dates, and Horoscope Insights of th...
 
How to Implement a Strategy: Transform Your Strategy with BSC Designer's Comp...
How to Implement a Strategy: Transform Your Strategy with BSC Designer's Comp...How to Implement a Strategy: Transform Your Strategy with BSC Designer's Comp...
How to Implement a Strategy: Transform Your Strategy with BSC Designer's Comp...
 
Unveiling the Dynamic Personalities, Key Dates, and Horoscope Insights: Gemin...
Unveiling the Dynamic Personalities, Key Dates, and Horoscope Insights: Gemin...Unveiling the Dynamic Personalities, Key Dates, and Horoscope Insights: Gemin...
Unveiling the Dynamic Personalities, Key Dates, and Horoscope Insights: Gemin...
 
amptalk_RecruitingDeck_english_2024.06.05
amptalk_RecruitingDeck_english_2024.06.05amptalk_RecruitingDeck_english_2024.06.05
amptalk_RecruitingDeck_english_2024.06.05
 
Zodiac Signs and Food Preferences_ What Your Sign Says About Your Taste
Zodiac Signs and Food Preferences_ What Your Sign Says About Your TasteZodiac Signs and Food Preferences_ What Your Sign Says About Your Taste
Zodiac Signs and Food Preferences_ What Your Sign Says About Your Taste
 
-- June 2024 is National Volunteer Month --
-- June 2024 is National Volunteer Month ---- June 2024 is National Volunteer Month --
-- June 2024 is National Volunteer Month --
 
Building Your Employer Brand with Social Media
Building Your Employer Brand with Social MediaBuilding Your Employer Brand with Social Media
Building Your Employer Brand with Social Media
 
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
How are Lilac French Bulldogs Beauty Charming the World and Capturing Hearts....
 
2024-6-01-IMPACTSilver-Corp-Presentation.pdf
2024-6-01-IMPACTSilver-Corp-Presentation.pdf2024-6-01-IMPACTSilver-Corp-Presentation.pdf
2024-6-01-IMPACTSilver-Corp-Presentation.pdf
 
Digital Transformation Frameworks: Driving Digital Excellence
Digital Transformation Frameworks: Driving Digital ExcellenceDigital Transformation Frameworks: Driving Digital Excellence
Digital Transformation Frameworks: Driving Digital Excellence
 
Best Forex Brokers Comparison in INDIA 2024
Best Forex Brokers Comparison in INDIA 2024Best Forex Brokers Comparison in INDIA 2024
Best Forex Brokers Comparison in INDIA 2024
 
How MJ Global Leads the Packaging Industry.pdf
How MJ Global Leads the Packaging Industry.pdfHow MJ Global Leads the Packaging Industry.pdf
How MJ Global Leads the Packaging Industry.pdf
 
Digital Marketing with a Focus on Sustainability
Digital Marketing with a Focus on SustainabilityDigital Marketing with a Focus on Sustainability
Digital Marketing with a Focus on Sustainability
 
The APCO Geopolitical Radar - Q3 2024 The Global Operating Environment for Bu...
The APCO Geopolitical Radar - Q3 2024 The Global Operating Environment for Bu...The APCO Geopolitical Radar - Q3 2024 The Global Operating Environment for Bu...
The APCO Geopolitical Radar - Q3 2024 The Global Operating Environment for Bu...
 
Hamster Kombat' Telegram Game Surpasses 100 Million Players—Token Release Sch...
Hamster Kombat' Telegram Game Surpasses 100 Million Players—Token Release Sch...Hamster Kombat' Telegram Game Surpasses 100 Million Players—Token Release Sch...
Hamster Kombat' Telegram Game Surpasses 100 Million Players—Token Release Sch...
 
Anny Serafina Love - Letter of Recommendation by Kellen Harkins, MS.
Anny Serafina Love - Letter of Recommendation by Kellen Harkins, MS.Anny Serafina Love - Letter of Recommendation by Kellen Harkins, MS.
Anny Serafina Love - Letter of Recommendation by Kellen Harkins, MS.
 
Business storytelling: key ingredients to a story
Business storytelling: key ingredients to a storyBusiness storytelling: key ingredients to a story
Business storytelling: key ingredients to a story
 
The 10 Most Influential Leaders Guiding Corporate Evolution, 2024.pdf
The 10 Most Influential Leaders Guiding Corporate Evolution, 2024.pdfThe 10 Most Influential Leaders Guiding Corporate Evolution, 2024.pdf
The 10 Most Influential Leaders Guiding Corporate Evolution, 2024.pdf
 

WEKA:Credibility Evaluating Whats Been Learned

  • 2. Training and Testing We measure the success of a classification procedure by using error rates (or equivalent success rates) Measuring success rate using training set is highly optimistic The error rate on training set is called resubstitution error We have a separate test set for calculating success error Test set should be independent of the training set Also some time to improve our classification technique we use a validation set When we hold out some part of training set for testing (which is now not used for training), this process is called holdout procedure
  • 3. Predicting performance Expected success rate = 100 – error rate (If error rate is also in percentage) We want the true success rate Calculation of true success rate Suppose we have expected success rate(f) = s/n, where s is the number of success out of a total n instances For large value of n, f follows a normal distribution Now we will predict the true success rate (p) based on the confidence percentage we want For example say our f = 75%, then p will lie in [73.2%,76.7%] with 80% confidence
  • 4. Predicting performance Now using properties of statistics we know that the mean of f is p and the variance is p(1-p)/n To use normal distribution we will have to make the mean of f = 0 and standard deviation = 1 So suppose our confidence = c% and we want to calculate p We will use the two tailed property of normal distribution And also that the are covered by normal distribution is taken as 100% so the are we will leave is 100 - c
  • 5. Predicting performance Finally after all the manipulations we have ,true success rate as: Here, p -> true success rate f - > expected success rate N -> Number of instances Z -> Factor derived from a normal distribution table using the 100-c measure
  • 6. Cross validation We use cross validation when amount of data is small and we need to have independent training and test set from it It is important that each class is represented in its actual proportions in the training and test set: Stratification An important cross validation technique is stratified 10 fold cross validation, where the instance set is divided into 10 folds We have 10 iterations with taking a different single fold for testing and the rest 9 folds for training, averaging the error of the 10 iterations Problem: Computationally intensive
  • 7. Other estimates Leave-one-out:Steps One instance is left for testing and the rest are used for training This is iterated for all the instances and the errors are averaged Leave-one-out:Advantage We use larger training sets Leave-one-out:Disadvantage Computationally intensive Cannot be stratified
  • 8. Other estimates 0.632 Bootstrap Dataset of n samples is sampled n times, with replacements, to give another dataset with n instances There will be some repeated instances in the second set Here error is defined as: e = 0.632x(error in test instances) + 0.368x(error in training instances)
  • 9. Comparing data mining methods Till now we were dealing with performance prediction Now we will look at methods to compare algorithms, to see which one did better We cant directly use Error rate to predict which algorithm is better as the error rate might have been calculated on different data sets So to compare algorithms we need some statistical tests We use Student’s t- test to do this. This test help us to figure out if the mean error of two algorithm are different or not for a given confidence level
  • 10. Comparing data mining methods We will use paired t-test which is a slight modification of student’s t-test Paired t-test Suppose we have unlimited data, do the following: Find k data sets from the unlimited data we have Use cross validation with each technique to get the respective outcomes: x1, x2, x3,….,xk and y1,y2,y3,……,yk mx = mean of x values and similarly my di = xi – yi Using t-statistic:
  • 11. Comparing data mining methods Based on the value of k we get a degree of freedom, which enables us to figure out a z for a particular confidence value If t <= (-z) or t >= (z) then, the two means differ significantly In case t = 0 then they don’t differ, we call this null hypothesis
  • 12. Predicting Probabilities Till now we were considering a scheme which when applied, results in either a correct or an incorrect prediction. This is called 0 – loss function Now we will deal with the success incase of algorithms that outputs probability distribution for e.g. Naïve Bayes
  • 13. Predicting Probabilities Quadratic loss function: For a single instance there are k out comes or classes Probability vector: p1,p2,….,pk The actual out come vector is: a1,a2,a3,…..ak (where the actual outcome will be 1, rest all 0) We have to minimize the quadratic loss function given by: The minimum will be achieved when the probability vector is the true probability vector
  • 14. Predicting Probabilities Informational loss function: Given by: –log(pi) Minimum is again reached at true probabilities Differences between Quadratic loss and Informational loss While quadratic loss takes all probabilities under consideration, Informational loss is based only on the class probability While quadratic loss is bounded as its maximum output is 2, Informational loss is unbounded as it can output values up to infinity
  • 15. Counting the cost Different outcomes might have different cost For example in loan decision, the cost of lending to a defaulter is far greater that the lost-business cost of refusing a loan to a non defaulter Suppose we have two class prediction. Outcomes can be:
  • 16. Counting the cost True positive rate: TP/(TP+FN) False positive rate: FP/(FP+TN) Overall success rate: Number of correct classification / Total Number of classification Error rate = 1 – success rate In multiclass case we have a confusion matrix like (actual and a random one):
  • 17. Counting the cost These are the actual and the random outcome of a three class problem The diagonal represents the successful cases Kappa statistic = (D-observed - D-actual) / (D-perfect - D-actual) Here kappa statistic = (140 – 82)/(200-82) = 49.2% Kappa is used to measure the agreement between predicted and observed categorizations of a dataset, while correcting for agreements that occurs by chance Does not take cost into account
  • 18. Classification with costs Example Cost matrices (just gives us the number of errors): Success rate is measured by average cost per prediction We try to minimize the costs Expected costs: dot products of vectors of class probabilities and appropriate column in cost matrix
  • 19. Classification with costs Steps to take cost into consideration while testing: First use a learning method to get the probability vector (like Naïve Bayes) Now multiple the probability vector to each column of a cost matrix one by one so as to get the cost for each class/column Select the class with the minimum(or maximum!!) cost
  • 20. Cost sensitive learning Till now we included the cost factor during evaluation We will incorporate costs into the learning phase of a method We can change the ratio of instances in the training set so as to take care of costs For example we can do replication of a instances of particular class so that our learning method will give us a model with less errors of that class
  • 21. Lift Charts In practice, costs are rarely known In marketing terminology the response rate is referred to as the lift factor We compare probable scenarios to make decisions A lift chart allows visual comparison Example: promotional mail out to 1,000,000 households Mail to all: 0.1%response (1000) Some data mining tool identifies subset of 100, 000 of which 0.4% respond (400) A lift of 4
  • 22. Lift Charts Steps to calculate lift factor: We decide a sample size Now we arrange our data in decreasing order of the predicted probability of a class (the one which we will base our lift factor on: positive class) We calculate: Sample success proportion = Number of positive instances / Sample size Lift factor = Sample success proportion / Data success proportion We calculate lift factor for different sample size to get Lift Charts
  • 23. Lift Charts A hypothetical lift chart
  • 24. Lift Charts In the lift chart we will like to stay towards the upper left corner The diagonal line is the curve for random samples without using sorted data Any good selection will keep the lift curve above the diagonal
  • 25. ROC Curves Stands for receiver operating characteristic Difference to lift charts: Y axis showspercentage of true positive X axis shows percentage of false positives in samples ROC is a jagged curve It can be smoothened out by cross validation
  • 26. ROC Curves A ROC curve
  • 27. ROC Curves Ways to generate cost curves (Consider the previous diagram for reference) First way: Get the probability distribution over different folds of data Sort the data in decreasing order of the probability of yes class Select a point on X-axis and for that number of no, get the number of yes for each probability distribution Average the number of yes from all the folds and plot it
  • 28. ROC Curves Second way: Get the probability distribution over different folds of data Sort the data in decreasing order of the probability of yes class Select a point on X-axis and for that number of no, get the number of yes for each probability distribution Plot a ROC for each fold individually Average all the ROCs
  • 29. ROC Curves ROC curves for two schemes
  • 30. ROC Curves In the previous ROC curves: For a small, focused sample, use method A For a large one, use method B In between, choose between A and B with appropriate probabilities
  • 31. Recall – precision curves In case of a search query: Recall = number of documents retrieved that are relevant / total number of documents that are relevant Precision = number of documents retrieved that are relevant / total number of documents that are retrieved
  • 32. A summary Different measures used to evaluate the false positive versus the false negative tradeoff
  • 33. Cost curves Cost curves plot expected costs directly Example for case with uniform costs (i.e. error):
  • 34. Cost curves Example with costs:
  • 35. Cost curves C[+|-] is the cost of predicting + when the instance is – C[-|+] is the cost of predicting - when the instance is +
  • 36. Minimum Description Length Principle The description length is defined as: Space required to describe a theory + space required to describe the theory’s mistakes Theory = Classifier and mistakes = errors on the training data We try to minimize the description length MDL theory is the one that compresses the data the most. I.e to compress a data set we generate a model and then store the model and its mistakes We need to compute: Size of the model Space needed to encode the error
  • 37. Minimum Description Length Principle The 2nd one is easy. Just use informational loss function For 1st we need a method to encode the model L[T] = “length” of the theory L[E|T] = training set encoded wrt the theory
  • 38. Minimum Description Length Principle MDL and clustering Description length of theory: bits needed to encode the clusters. E.g. cluster centers Description length of data given theory: encode cluster membership and position relative to cluster. E.g. distance to cluster centers Works if coding scheme uses less code space for small numbers than for large ones
  • 39. Visit more self help tutorials Pick a tutorial of your choice and browse through it at your own pace. The tutorials section is free, self-guiding and will not involve any additional support. Visit us at www.dataminingtools.net