Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Ahmet Selman Bozkır, Ebru Akçapınar SezerAhmet Selman Bozkır, Ebru Akçapınar SezerHacettepe University Computer Engineerin...
 The importance of subject What is “Data Mining” ? Microsoft Decision Trees Data attributes Model building The tool ...
 Economics is the social science that analyzes theproduction, distribution, and consumption of goodsand services. (Wikipe...
Data mining: is the process of extracting hidden & useful information from raw databy utilizing statistics, AI and machine...
Predictive Methods1. Classification (DecisionTrees, Bayesian Classification, etc…)2. Regression (CART, Kernel Ridge Regres...
 •Invented by MS, in 1999•Designed for both classification and regression•Works on both categorical & numerical attribute...
General dataVariable name Type UsageDay Continuous Input Day number ranging from 1 to 31Month Continuous Input Month numbe...
Overall data is 44 months period between 1.1.2008 – 21.8.2011- 1323 recordsValidation set (92)Train + Test Data (1231)1323...
120 different MSDT modelscovering Students,Academics,Officers data“0.4” “0.5” “0.7” and “0.85” complexity_panalty values a...
 Purely web based application Microsoft Analysis Services ASP.NET technology ADOMD.NET andDMX queries Platform indepe...
 Food consumption prediction is important due to balance between supplyand demand. Decision trees is a fine candidate fo...
 Questions?
ADEM: An Online Decision Tree Based Menu Demand Prediction Tool for Food Courts
ADEM: An Online Decision Tree Based Menu Demand Prediction Tool for Food Courts
Upcoming SlideShare
Loading in …5
×

ADEM: An Online Decision Tree Based Menu Demand Prediction Tool for Food Courts

824 views

Published on

Published in: Business, Technology, Education
  • Be the first to comment

  • Be the first to like this

ADEM: An Online Decision Tree Based Menu Demand Prediction Tool for Food Courts

  1. 1. Ahmet Selman Bozkır, Ebru Akçapınar SezerAhmet Selman Bozkır, Ebru Akçapınar SezerHacettepe University Computer Engineering Dept.Hacettepe University Computer Engineering Dept.Ankara -TurkeyAnkara -Turkey20.5.201320.5.2013
  2. 2.  The importance of subject What is “Data Mining” ? Microsoft Decision Trees Data attributes Model building The tool ADEM Conslusion & Discussion
  3. 3.  Economics is the social science that analyzes theproduction, distribution, and consumption of goodsand services. (Wikipedia)
  4. 4. Data mining: is the process of extracting hidden & useful information from raw databy utilizing statistics, AI and machine learning techniques and smart algorithms.
  5. 5. Predictive Methods1. Classification (DecisionTrees, Bayesian Classification, etc…)2. Regression (CART, Kernel Ridge Regression etc..)3. Artificial Neural Networks4. Kernel Based Methods (SVM, RVM, Gaussian Processes etc...)Descriptive Methods1. Clustering (K-Means , Hierarchical Clustering, EM etc…)2. Association Rules (Apriori, GRI etc..)Decision Trees SVM ANN
  6. 6.  •Invented by MS, in 1999•Designed for both classification and regression•Works on both categorical & numerical attributes•Serves entropy, Bayesian K2, and Bayesian DirichletEquivalent with Uniform prior choices as splitting criteria•Uses multi-way splits and support binary splitting•Missing value handling is provided•Avoids tree pruninginstead it uses complexity_penalty parameter to control thedepth of tree
  7. 7. General dataVariable name Type UsageDay Continuous Input Day number ranging from 1 to 31Month Continuous Input Month number ranging from 1 to 12Day Name Discrete Input The name of the day ranging from Monday to SundayIs Holiday Boolean Input A flag variable (0/1) denoting that day is in weekend/holidayor weekdayCalorie Continuous Input Total calorie amount of the menuisMeatMenu Boolean Input Meat / Vegetarisch (True/False)containsDessert Boolean Input Contains Dessert (True/False)Food1 Discrete Input First food name in menuFood2 Discrete Input Second food name in menuFood3 Discrete Input Third food name in menuFood4 Discrete Input Fourth food name in menuSales-Student-Lunch Continuous Predict Number of sales for students in lunch sessionSales-Academic-Lunch Continuous Predict Number of sales for academic staff in lunch sessionSales-Officials-Lunch Continuous Predict Number of sales for officers in lunch session
  8. 8. Overall data is 44 months period between 1.1.2008 – 21.8.2011- 1323 recordsValidation set (92)Train + Test Data (1231)1323 records(All data)10 fold cross validationfor robust model selection10 models for eachcustomer type (3) with 4 CPparameter = 120 models
  9. 9. 120 different MSDT modelscovering Students,Academics,Officers data“0.4” “0.5” “0.7” and “0.85” complexity_panalty values andVAF values are compared -> best : 0,87 0.5 CP value and the 7th fold training data selected as thebase predictor model dataComplexity Penalty Value # of times it is winner0.4 30.5 60.7 20.85 1
  10. 10.  Purely web based application Microsoft Analysis Services ASP.NET technology ADOMD.NET andDMX queries Platform independentDecision Tree ModelsADOMD.NETClientMicrosoft Analysis ServicesMiningEngineAuthenticationADEMOnline QueryTree ExplorerVisualizationIIS 7.0 / ASP.NET 2.0
  11. 11.  Food consumption prediction is important due to balance between supplyand demand. Decision trees is a fine candidate for revealing food court consumptionpatterns Multi-way splitted decision trees have model transparency which has themodel explanation power More data improve the results however seasonal fluctuations should becarefully traced. Vegetarish selections are important on consumption amount in our casestudy. A decision tree powered web-based decision support system is feasible andvaluable for group decision making
  12. 12.  Questions?

×