logistic-regression

Artificial Intelligence and Computer Graphics - First Assignment

Problem

The dataset in bank −marketing−campaign.zip represents a collection of examples from a marketing campaign organised by a bank to get its clients to place a term deposit. The dataset has 21 columns, described in Table 1. Your task is to build a classifier based on this dataset. The classifier should use the regularised logistic regression algorithm, with the regularised cross-entropy as its cost function. You will use the sum of the magnitude of all the coefficients (also known as Lasso regularisation or L 1 regularisation) as your regularisation technique.

Assessment Criteria

The following criteria will be followed to assess your submission:

• Data cleaning and preparation in Julia;

• Implementation (from scratch) of the regularised logistic regression al- gorithm in Julia;

• Design and implementation of the classifier;

• Performance metrics including (It is advised to use a confusion matrix.):

  accuracy: the proportion of correct predictions (clients correctly pre-dicted to have placed a term deposit or not) over all predictions;

  precision: the proportion of clients the classifier predicted have placed a term deposit actually did so;

  recall: the proportion of clients that actually placed a term deposit which was predicted by the classifier

  

Submission Instructions

• This project is to be completed by groups of maximum two (2) students each.
• For each group, a repository should be created either on Github or Gitlab. The URL of the repository should be communicated by Thursday, May 14 th 2020, with all group members set up as contributors.
• The submission date is Monday, May 25 th 2020, midnight.
• A submission will be assessed based on the clone of its repository at the deadline.
• Any group who fails to submit on time will be awarded the mark 0.
• There should be no assumption about the execution environment of your code. It could be run using a specific framework or simply on the command line.
• In the case of plagiarism (groups copying from each other or submissions copied from the Internet), all submissions involved will be awarded the mark 0, and each student will receive a warning.