Assignment One

This assignment expects you to make a use of multiple machine learning algorithms

to make predictions from the following datasets. Refer to our lecture notes and

practicals to use appropriate algorithms as per the outline defined.

Datasets

The dataset "film$\_$collection$\_$dataset.csv$"$ contains information about movies and

their marketing, production expense, budget of the movie, length of the movie, critic

rating etc and the money earned.

The dataset "loan$\_$dataset.csv$"$ contains people's personal information and a

classification field "loan$\_$status" states whether or not their request to loan was

approved based on their education, income and credit score.

The dataset "marketing$\_$campaign$\_$dataset.csv$"$ contains data about people's

education, marital status, income, number of kids in the household etc and their

preferences to multiple products and their binary response $($acceptance$/$rejection$)$ to

multiple offers made in campaigns $($from columns AcceptedCmp$1$ to AcceptedCmp$2$

and the response column$).$ The dataset also contains information about the amount of

money spent on products such as Gold, Fruits, Meat, Fish, Sweets and Wines in the

last two years.

Outline

Create optimum training$/$testing split to form appropriate machine learning

models for both classification and regression problems and also make a use of

cross validation methods to avoid model overfitting problems.

Achieve necessary data pre$-$processing steps including outlier removals and

appropriate visualisation steps such as pair plots or correlation matrix to better

understand the data distribution.

Create Linear and Multiple regression models to predict the revenue of movies

by proposing unseen input data by keeping in mind the concept of

multicollinearity. Also calculate the coefficient of determination squared$).$ Perform the analysis with and without data standardisation to

differentiate the prediction effect.

Make a decision tree model $($for a regression problem$)$ using optimum training$/$

testing split and calculate the Mean Squared Error $($MSE$)$ to check the model's

accuracy. Also predict some unseen movies data and compare the model's

accuracy against the regression model to find out which model performs better.

Train the Logistic Regression and Decision Tree models with optimum

train$/$test split for solving classification problems using GridSearch

Hyperparameter tuning to predict whether or not a loan of a certain profiles of

individuals would be approved. Also employ the Random Forest classification

model to predict the class of the same unseen data $($calculating the accuracy of

the model$)$ and compare the results with the Logistic Regression and Decision

Tree models and evaluate your analysis.

Use the same classification models for the Marketing campaign dataset and

predict whether individual profiles with certain characteristics $($such as marital

status, income or education level$)$ is likely to respond to the campaigns made.

Also use the K$-$means clustering algorithm to identify clusters of people with

certain characteristics $($such as education, marital status or income level$)$ and

the money they spent on products like Gold, Fruits, Meat, Fish, Sweets and

Wines etc.

In your report, show appropriate visualisations, confusion matrix and

classification report for each classification model wherever necessary.

Note: Students are allowed to structure the report as they find appropriate. Please use

the IEEE referencing style in your work to cite appropriate sources.