Assignment 1

This folder contains the starter files for Assignment 1, which is about solving a classification problem.

Submission is done by pushing the necessary files to github, as detailed below. It will be possible from Monday, once a private github repository for each team has been created.

Data

The assignment uses the Adult dataset. The task is to predict, based on census data, whether the income of an individual exceeds $50K/yr.

• Training set: data/adult.data. The file contains 32561 records in total. Each line corresponds to a record, with the attributes comma separated. The last attribute is the target class label: <=50K or >50K.
• Test set: data/adult.test. The file contains 16281 records. It has the exact same format as adult.train, except the last column. I.e., the class labels are missing from the test file; your task is to predict these and output the predictions to a separate file.
• There are 14 attributes, a mixture of categorical and continuous ones. Mind that there are missing attribute values (denoted by ?). See the adult.names file under the data folder for further details.
• The eval.py script can be used for evaluation during development (if the holdout method or cross-validation is employed).
  • The toy_data folder contains a toy-sized ground truth set and predictions; this is only provided to allow you to try out the evaluation script. Run python eval.py toy_data/test.gt toy_data/test.pred from the assignment's root folder.

Task 1

Predict the target class labels for the test data using a decision tree classifier that you implement from scratch.

• For each record in the test data, you need to output one line with the target label (either <=50K or >50K).
• The output must be placed in the output/task1.out file. If the file exists, it will be automatically evaluated upon git push.
• Implementing from scratch means that you are not allowed to use an existing library or package that implements decision trees.
• You are free to pick the programming language/environment, but you are required to submit the complete source code. The code used needs to be placed in the code folder.
• The output file must be generated automatically using the decision tree approach implemented by you. Submitting predictions from any other source (Internet, another team, etc.) is considered cheating and will result in immediate disqualification (i.e., dismissal from the course).
• To have the assignment approved, you will need to reach an Accuracy of at least 0.8 on the test set.
• Deadline: Sep 28, 23:59

Task 2

Perform some data mining on the Adult dataset using the decision tree you built and write a short report describing the process and your findings.

• The report is min 3, max 8 pages (A4) long, written in English.
  • The specific format is up to you, max font size is 12pt.
• It needs to contain the followings:
  • Instructions on how to run your code
  • Report on any data exploration you performed (e.g., summary statistics and/or visualizations of attributes)
  • Data processing steps applied (e.g., variable transformations, feature creation)
  • How you dealt with missing attribute values
  • Looking at the first 3 levels of the decision tree, what observations can you make? Which attributes are the most important?
• The document must be in pdf format and placed under the report folder.
• Deadline: Oct 5, 23:59

Task 3

This is the same as Task 1, but any existing machine learning tool/library/package may be used.

• This task is optional.
• The same rules apply as for Task 1, regarding supplying all source code and cheating.
• You can submit the approach and output that you developed for Task 1.
• The output must be placed in the output/task3.out file. If the file exists, it will be automatically evaluated upon git push.
• Deadline: Sep 28, 23:59

Leaderboard

The URL to the online leaderboard will be provided here.

Results are considered in two categories:

• Decision tree track -- this is for everyone, based on submissions to Task 1
• Open track -- this is optional, based on submissions to Task 3

The best performing team for each track gets 5 bonus points at the exam (all members).