SEEM4630 E-Commerce Data Mining
2024-2025 Second Term
Assignment 3
Due: April 11 17:00
Problem Statement
This assignment involves using Weka to predict client subscription to term deposits in a
bank. The prediction process consists of three phases:
1) data preprocessing,
2) modeling and model assessment,
3) utilization of the learned model for as prediction for new clients.
This assignment involves all three phases. In fact, the process is somewhat similar to Assignment 1
and Assignment 2 with some modifications and enrichment. The details are given below.

Requirements
This is an individual assignment, and you must follow the requirements listed below.
1. Use the "Bank Marketing (with social/economic context)" dataset, specifically the "bank-
additional.csv" file, to build the model.
2. The target variable is "y" which indicates whether the client subscribed to a term deposit or not.
3. Similar to the spirit of Assignment 1, you will conduct data pre-processing. However, you need
not follow the precise data preprocessing as in Assignment 1. Instead, you can perform any
kind of data pre-processing. Also, for data preprocessing, you can use any other software
packages (not necessarily Weka) or write your own programs.
4. Split the whole dataset into two subsets, namely, a training set and a testing set.
5. Save the training and testing sets as two separate files for subsequent processing. Note that the
format of the files needs to be Weka-compatible, i.e. arff format.
6. Similar to the spirit of Assignment 2, you use the training set to train a model and use the testing
set to measure the performance of the model.
7. Repeat the above steps as many times as you like. However, you can only use any data mining
/ machine learning models provided in Weka, e.g. decision tree model, neural network model,
or logistic regression model. You can try any parameters in these models. Note that only data
mining / machine learning models in Weka are allowed.
8. After some trials, you should decide and save the most desirable model for delivering to the
bank for operation.
9. Similar to Assignment 2, the delivered model will be used for operation and the model will
predict the term deposit subscription of new clients. To simulate this operational setting for a
bunch of clients, we provide a new client data set, namely, bank-new-client.csv. Note that the
true class labels in this data set are hidden. Your aim is to make prediction for each of the new
clients. Since the data in this new client data set was not available during the learning stage,
this new data has not been pre-processed. Therefore, in order to make it suitable for use, you
should perform appropriate data pre-processing. Note that the format of the files needs to be
Weka-compatible, i.e. arff format.
10. Load the pre-processed model obtained in Step 8. Then use it to conduct prediction on the pre-
processed file in Step 9. Output the prediction results to a CSV file similar to the format
specified in Assignment 2.
You need to write a report describing the process of developing the delivered model, including data
pre-processing, selected model and model parameters. You will submit the report in PDF format.
You should write clearly the process. In addition, for the delivered model, you should also provide
Weka screenshots demonstrating the data pre-processing, the adoption of the data mining model and
its parameters, the accuracy of the testing data, and the prediction results of new clients.
Assessment Criteria
This assignment will be assessed based on a range of assessment components including quality of
new client prediction, rationale and quality of the whole process, quality of report presentation, etc.
The component related to the quality of new client prediction will attain a high weighting. For this
assessment component, the TA will measure the prediction accuracy and rank the accuracy among
students. The higher the rank, the higher of this assessment component score.

Assignment Submission
1. There are four files for this assignment to be submitted:
a) One model object file representing the delivered model;
b) One CSV file for the prediction results;
c) One PDF file for the report;
d) One ARFF file for the pre-processed new client data set.
2. Compress these four files into a single ZIP file. Ensure all the file names follow the "id-name"
format. For example, if your name is Chan Tai Ming and your student ID is 00123456, the files
should be named as follows:
a) “00123456-CHANTaiMing-Assignment3-model” for the model object file;
b) “00123456-CHANTaiMing-Assignment3-predict.csv” for the prediction results;
c) “00123456-CHANTaiMing-Assignment3-report.pdf” for the assignment report;
d) “00123456-CHANTaiMing-Assignment3-input.arff” for the pre-processed new
client data set;
e) “00123456-CHANTaiMing-Assignment3.zip” for the ZIP file.
3. Upload the ZIP file to Blackboard for submission.
4. You may resubmit your assignment before the deadline, with only the latest submission being
assessed.
5. For any inquiries, please email xut0092@link.cuhk.edu.hk

学霸联盟