INT104 Coursework 2 Solution

Introduction
A spreadsheet of mark is provided where the mark for each question is listed for each student. There are more than 500 students in total who comes from 4 majors. The majors in the spreadsheet are labelled as ‘1’, ‘2’, ‘3’ and ‘4’. There are also a few students that belong to other majors, which are labelled as ‘0’. The students whose major is labelled ‘0’ could be treated either as outlier data or could be simply deleted. This coursework requires the student to design a classifier to classify students to the program they belonged to, according to the mark earned for each question.


Though Python is recommended to be used as the implementation tool for this coursework, other tools / programming language such as Microsoft Excel, Weka, MATLAB, Java and C++ are also acceptable. The student could use MULTIPLE types of tools throughout three tasks in this coursework. The lab report should be submitted in PDF format, which could be produced by any typesetting software such as Microsoft Word, Latex (recommended) and Markdown.

Task 1: Data observation
With the presented data, the student is required to make reasonable data observation in this task. The results of data observation could be presented as data visualisation. As we have learned from the lectures, all data contains bias that is caused by different reasons. The bias of some features in data could lead to a good classifier to a dedicated machine learning task. This task requires the student to find their own ways to show the data bias and encourages the student to find data bias that is closely related to the given labels. Despite commonly used, data features such as average value, median value, range and standard deviations are not necessary to be the features showing data bias. Data features resulted from more complex mathematical operations such as Principal Component Analysis (PCA), Discrete Fourier Transform (DCT) and Non-negative Matrix Factorisation (NMF). For different features, there might be a bias of distribution hence a process of removing bias in features should also be considered and presented.

In this task, the student should try different ways to extract data features of the given dataset and decide which data feature will be used as the input of classifier that to be developed in the next task. In the lab report, a full and detailed justification on feature selection (including bias removal process) should be presented with no less than five features as candidates.
Task 2: Build Classifiers

The candidate classifier should be built in a supervised way. The classifier could be the method taught in the lectures such as Support Vector Machine (SVM) and Decision Tree. However, the student is also encouraged to try methods beyond the scope of the lectures delivered such as deep neural networks and Bayesian graphical models.

In the lab report, the student should specify what method is used to build the candidate classifiers, what data feature is used as the input of the candidate system and what result is obtained. The key part of program such as data pre-processing, the training process of the model and the inference process of the system should be stated with details. However, the use of screenshot should be avoided, where a text-based description (such as text-based source code with line number) is preferred when necessary. (NOTE: source code presented in the lab report should be in Courier New or similar fonts.)

The student is then required to recommend ONE classifier among the three candidate classifiers. The process of classifier evaluation should be fully demonstrated (e.g., cross validation process). The decision of recommendation should also be fully justified. The process of classifier evaluation and the justification of classifier choice should be documented in the lab report.
Task 3: Unsupervised Classification for Student Classification
In this task, the student is required to classify students to different groups in an unsupervised manner, according to marks awarded for each exam question. The classification of groups should make the delivery of lecturers in the next semester easier by classifying students in the same group sharing comparable properties of studying. There are no dedicated ways of classification hence the student should make their choice of principle applied to classification (i.e., there are no dedicated number of groups that the student should be classified into).

In the lab report, the student is expected to present the full details of the unsupervised classification process and fully justify the final decision made for student classification. Specifically, the student should interpret the principle that is followed for the classification. As the case of earlier tasks, screenshots of source code should be avoided.
Lab Report
Please note, a title should also be included in the lab report. (Please do NOT use “lab report” as the title of the report.). The page limit for the lab report is 8 pages excluding abstract, reference list, cover page and appendix. The majority of marks in this coursework will be awarded according to the content of lab report. Please be aware that a lengthy report
does not guarantee a high mark for this coursework. It is not necessary to attach the source
code in the lab report.

A separate lab report template will be provided as a reference. However, the outline provided need not to be strictly followed.

Marking criteria
Marking Criteria Item Marks Description
Editorial and
Language
Issues
(20) Formatting 5 Format should follow IEEE-like style with double columns:  Font size: 9 - 11 pts
 Line spacing: 1
 Alignment: Justified
 Font: Times New Roman
 Margins: Conventional
Organization 5 The paper should be well organised showing a clear structure of the report as instructed in this material.
Spelling 5 The spelling in the paper should be completely correct. Any incorrect spelling can result in a loss of marks.
Grammar 5 The language used in the essay must be academic language.
Task 1
(20) Feature Extraction 10 Showing reasonable amount of feature extraction attempts.
Data
Visualisation 10 Visualise the extracted features of raw data to show the bias of distribution.
Task 2
(30) Description of
Classification Methods 10 Describe the selected methods of classifications with justification.
Training
Classifiers 10 Training at least three classifiers (no more than five) and obtain predictions.
Classifier Selection 10 Using the obtained classifiers to classify given data and evaluate performance of the candidates.
Task 3
(20) Unsupervised
Classification 10 Classify students into several groups via one unsupervised classification method.
Critical
Thinking in
Interpretation 10 The students should interpret the classification results with a reasonable interpretation of language.
Lab
Sessions
(10) Time
Missing all live demonstrations is likely to result to a mark of 0.
Submission
1. Only submissions in PDF format are accepted.
2. Submit your lab report via the dedicated Learning Mall coursework link.
3. Please name your submission file as ID_FirstName_LastName_C2.pdf (e.g., 1234567_FirstName_Surname_C2.pdf).
Resit
In the case of resitting this coursework, the student should follow the same instructions as in this document. However, the resitting task of this coursework should use a separate dataset
provided. The experiment report should be submitted in the dedicated way as instructed by separate mails.