This document describes an author paper identification problem where the goal is to determine the correct author for a given paper from a dataset of author information. It discusses preprocessing the data to clean issues and extract relevant features. Random forest and gradient boost models are built and evaluated on test data to solve the problem. Key steps taken include data cleaning, feature engineering from the paper, author and paper-author data, model building using Weka, Mahout and H2O, and evaluating the results using mean average precision.

1. Author- Paper Identification Problem
Guided By
Prof Duc Tran
Team :
Karthik Reddy Vakati
Nachammai C
Pooja Mishra

2. Problem Statement
To determine the correct author from the author’s dataset
for a particular paper.
Ambiguity in author names might cause a paper to be
assigned to the wrong author, which leads to noisy
author profiles.
Challenge is to determine which papers in an author
profile were truly written by a given author

3. Data Points
The data points include all papers written by an author, his
affiliation (University, Technical Society, Groups).
 Paper-Author -( PaperId , AuthorId, Name, Affiliation)
The meta data includes journals written by him and
conferences attended by an author.
 Paper -( Id, Title, Year, ConferenceId , JournalId,
Keywords)
 Author -( Id, Name, Affiliation)
 Conference-(Id, ShortName,FullName,HomePage)
 Journal -(Id, ShortName, FullName, HomePage)

4. Machine Learning Task
• Building the model using train dataset and testing
• Feature Engineering
• Algorithms
• Model Tuning
• Results
• Evaluation

5. Steps Taken to Solve Problem
 Data preprocessing and cleaning
 Feature engineering
 Extracting feature values
 Creating final input file
 Choose a ML algorithm - Random Forest/Gradient
Boost Model
 Building the model
 Tuning the model
 Test the model on test data
 Evaluating the results

6. Data preprocessing and cleaning
Issues with data
Few had attributes spilled over 3 rows
Some rows had more attributes than the required
number of attributes
Wrote a Perl script to Clean data and format it
Removed stop words using NLTK package in python
Converted all text to lower case
Removed special characters
Removed noise from years field
Assigned ID to each keyword and normalized it

7. Issues with data-I

8. Feature Engineering Steps
• Aggregation: combining multiple features into one.
How did we use: Elaborated train file with AuthorID, PaperID and
Confirmation combined with name and affiliation from Author file and
PaperAuthor file.
• Construction: Creating new features out of original ones
How did we use: minyear and maxyear into active years
• Discretization: Converting continuous features or variables to discretized
or nominal features
How did we use: The year the paper is published. The max and min years
the author was actively publishing papers.

9. Author features
• distance between the author names in paper-author and author files
• matched substring ratio between the author names in paper-author and author
files
• keywords used by a particular author(less weight)
• count keywords for author
• count the no of co-authors for a given co-author
• weighted TF-IDF measure of all author keywords inside author's papers
• count different papers of author
• years during which the author wrote many papers
• number of times an author is repeated (sum for distinct ids)
• list of distinct ids assigned to the same author

10. Paper features
• year of paper
• count authors of paper
• count duplicated papers for paper
• count duplicated authors for paper
• count keywords in paper
• how many time the exact same set of authors is repeated in
different papers (without
duplicates)

11. Paper-Author features
• correct affiliation from the table PaperAuthor: binary feature
• which year the author publishes (for the first year papers of author this
feature equals 1 for the second year papers - 2 and so on)
• count sources: number of times pair author-paper is appeared in the table
PaperAuthor table and Author table the same

12. Machine Learning Models Used
Models used earlier
• K-means clustering
• ZeroR
• Tree J-48
• Naïve Bayes

13. Machine Learning Models Used
• RandomForest
 Using Weka, Mahout and H20
• Gradient Boost
 Using H20

14. Build Random forest using weka
With the elaborated train file

15. Build Random forest using H20

16. Feature values extraction & importance
Count of paper for each author
Maximum active year for given author
Maximum active year for given author
Jaccard distance between author name in author file and
paper author file
Jaccard distance between affiliation in author file and paper
author file
The year paper was published
Normalized Keyword ids

17. 1. Put the data in HDFS.
$HADOOP_HOME/bin/hadoop fs -mkdir testdata $HADOOP_HOME/bin/hadoop fs -put testdata
2. Build the job files.
$MAHOUT_HOME/ run: mvn clean install -DskipTests
3. Generate a file descriptor for the dataset.
$HADOOP_HOME/bin/hadoop jar
$MAHOUT_HOME/core/target/mahout-core--job.jar org.apache.mahout.classifier.df.tools.Describe -p
testdata/KDDTrain+.arff -f testdata/KDDTrain+.info -d N 3 C 2 N C 4 N C 8 N 2 C 19 N L
4. Run the model
$HADOOP_HOME/bin/hadoop jar
$MAHOUT_HOME/examples/target/mahout-examples--job.jar
org.apache.mahout.classifier.df.mapreduce.BuildForest -Dmapred.max.split.size=1874231 -d
testdata/KDDTrain+.arff -ds testdata/KDDTrain+.info -sl 5 -p -t 100 -o nsl-forest
5. Using the Decision Forest to Classify new data.
$MAHOUT_HOME/examples/target/mahout-examples--job.jar
org.apache.mahout.classifier.df.mapreduce.TestForest -i nsl-kdd/KDDTest+.arff -ds nsl-kdd/KDDTrain+.info -m
nsl-forest -a -mr -o predictions
Random Forest using Mahout Steps

18. Evaluation Metrics
The mean average precision for N users at position n is
the average of the average precision of each user, i.e.,
MAP@n=∑i=1Nap@ni/N
• Mean Average Precision

19. Lessons Learned!
Since we were given train and test data supervised learning is the best fit. Hence
classification algorithms work better for author paper identification problem rather
than clustering.
When you want to find patterns or structures in the provided data use
unsupervised learning models such as clustering.
Choosing the features is the most important thing and we can extract the feature
values from the given data and use it to build the model.
Construct an initial set of features and try to build the model and test for its
accuracy. This might not give you better results always. Construct features from
features.
Choosing features from different data points will give better results than just
choosing them from only one.
Choose initial set of weights for each feature based on its importance. This will
help in model tuning.

20. Thank you!!