Assessment Proforma 2024-25

Key Information

The Assessment Calendar can be found under ‘Assessment & Feedback’ in the COMSC-ORG-SCHOOL organisation on Learning Central. This is the single point of truth for (a) the hand out date and time, (b) the hand in date and time, and (c) the feedback return date for all assessments.

Learning Outcomes

The learning outcomes for this assessment are as follows:

1.  Implement and evaluate machine learning methods to solve a given task

2.  Explain the fundamental principles underlying common machine learning methods

3.  Choose an appropriate machine learning method and data pre-processing strategy to address the needs of a given application setting

4.  Reflect on the importance of data representation for the success of machine learning methods

5.  Critically appraise the ethical implications and societal risks associated with the deployment of machine learning methods

6.  Explain the nature, strengths and limitations of an implemented machine learning technique to an audience of non-specialists

Submission Instructions

The coversheet can be found under ‘Assessment & Feedback’ in the COMSC-ORG- SCHOOL organisation on Learning Central.

All files should be submitted via Learning Central.  The submission page can be found under ‘Assessment & Feedback’ in the CMT316 module on Learning Central.  Your submission should consist of multiple files:

This coursework consists of a portfolio divided into two parts

-    Part (1) consists of selected homework similar to the one handed in throughout the course. The final deliverable consists of a single PDF file, which may include methodology, snippets of Python code and solved exercises.

-    Part (2) consists of a machine learning project where students implement a basic machine learning algorithm for solving a given task. The deliverable is a zip file with the code, and a written summary (up to 1200 words) describing solutions, design choices and a reflection on the main challenges faced during development.

Any code submitted will be run in Python 3 (Linux) and must be submitted as stipulated in the instructions above.

Any deviation from the submission instructions above (including the number and types of files submitted) may result in a mark of zero for the assessment or question part.

If you are unable to submit your work due to technical difficulties, please submit your work viae-mailto comsc-submissions@cardiff.ac.ukand notify the module leader.

Assessment Description

In this coursework, students demonstrate their familiarity with the topics covered in the module via two separate parts with the first part worth 40% and the second part worth  60%.

Part 1 (40%)

In Part 1, students are expected to answer two practical questions.

1.  Your algorithm gets the following results in a classification experiment. Please compute the precision, recall, f-measure and accuracy *manually* (without the help of your computer/Python, please provide all steps and formulas). Include the process to get to the final result. (15 points)

2.  You are given a dataset (named “real_estate”) with different house properties (dataset available in Learning Central). Your goal is to train machine learning models in the training set to predict the house price of a unit area in the test set. The problem should be framed as both regression and classification. For regression, the house price of a unit area is given; for classification, there would be two labels (expensive and not- expensive) depending on the house price of a unit area: expensive if it is higher or equal to 30, and not-expensive if it is lower than 30.

The task is therefore to train two machine learning models (one regression and another one classification) and check their performance. The student can choose the models to solve this problem. Write, for each of the models, the main Python instructions to train and predict the labels (one line each, no need to include any data preprocessing instructions in the pdf) and the performance in the test set in terms of Root Mean Squared Error (regression) and accuracy (classification). While you will need to write the full code to get to the results, only these instructions are required in the pdf. (25 points)

Part 2 (60%)

In Part 2, students are provided with a text classification dataset (named “bbc_news”). The dataset contains news articles split into five categories: tech, business, sport, politics and entertainment. Based on this dataset, students are asked to preprocess the data, select features and train and evaluate a machine learning model of their choice for classifying news articles. Students should include at least three different features to train their model, one of them should be based on some sort of word frequency. Students can decide the type of frequency (absolute or relative, normalized or not) and text preprocessing for this mandatory word frequency feature. The remaining two (or more) features can be chosen freely. Then, students are asked to perform. feature selection to reduce the dimensionality of all features.

Note: Training, development and test sets are not provided. It is up to the student to decide the evaluation protocol and partition (e.g., cross-validation or pre-defining a training, development and test set). This should be explained in the report.

Deliverables for this part are the Python code including all steps and a report of up to 1200 words. The Python code should include the Python scripts and a small README file with instructions on how to run the code in Linux. Jupyter notebooks with clear execution   paths are also accepted. The code should take the dataset set as input, and output the results according to the chosen evaluation protocol. The code will consist of 25% of the  marks for this part (15 points) and the report the remaining 75% (45 points). The code  should contain all necessary steps described above: to get the full marks for the code, it should work properly and clearly perform all required steps. The report should include:

1)  Description of all steps taken in the process (preprocessing, choice of features,

feature selection and training and testing of the model). This description should be such that one could understand all steps without looking at the code (15 points - The quality of the preprocessing, features and algorithm will not be considered here)

2)  Justification of all steps. Some justifications may be numerical, in that case a

development set can be included to perform. additional experiments. (10 points - A reasonable reasoned justification is enough to get half of the marks here. The usage of the development set is required to get full marks)

3)  Overall performance (accuracy, macro-averaged precision, macro-averaged recall and macro-averaged F1) of the trained model in the dataset. (10 points -

Indicating the results, even if very low, is enough to get half of the marks here. A minimum of 65% accuracy computed on a test set of at least 10% of the dataset is required to get full marks)

4)  Critical reflection of how the deliverable could be improved in the future and on   possible biases that the deployed machine learning may have. (10 points - The depth and correctness of insights related to your deliverable will be assessed)

The report may include tables and/or figures.

Assessment Criteria

Credit will be awarded against the following criteria. Credit will be awarded against the following criteria.

➢ Part 1. The main criteria for assessment is based on the correctness of the

answers, for which the process is also required. Full marks will be given for answers including both the correct answer and a correct justification or methodology.

➢ Part 2. This part is divided into Python code (25%) and an essay (75%). The code will be evaluated based on whether it works or not, and whether it minimally contains the necessary steps required for the completion of Part 2. Four items will be evaluated in the essay, whose weights and descriptions are indicated in the assessment instructions. The main criteria to evaluate those items will be the adequacy of the answer with respect to what was asked, and the justification provided.