Project 1 : Moodle Database: Educational Data Log Analysis

Focused on analyzing learners' activity in the Moodle Learning Management System, stored in a database, to understand the learners' skill and knowledge development.

Involves:

Moodle database schema understanding

• Understanding the moodle database schema by exploring data in the database
• Writing python class to pull data from the database.

Data Extraction Transformation and Loading (ETL)

• Using python class to perform log analysis

Dashboard making with Tableau

Project 2 : Bank Institution Term Deposit Predictive Model

Guided by a dataset from the bank of Portugal, this project aimed at developing a machine learning model to predict whetehr a customer would subscribe to a term deposit or not.

Involves:

Exploratory Data Analysis

• Understanding class imbalance
• Univariate and Bivariate Analysis of categorical and numerical features
• Correlation of numerical features

Data Cleaning

• checking for missing values
• checking and dropping duplicated rows
• Checking for and dealing with outliers in numerical features

Data Pre-processing

• One hot encoding of categorical features
• Stardardization of numerical features
• Dimensionality reduction using t-SNE, PCA and Autoencoders
• Feature selection using RFE
• Class balancing using SMOTE

Development of Robust Machine Learning Algorithms

Involved model training using:

• Multilayer Perceptron
• XGBoost
• Random Forest
• Decision Trees

Other concepts explored here are:

• K-fold cross-validation
• Performance evaluation using confusion matrix, precision, recall, F-measure and support

Sample t-SNE plot derived in the project:

Project 3 : Pharmaceuticals Sales Prediction API

A group project that aimed at developing a sales prediction API for multiple pharmaceutical stores. Involves:

Exploration of customer purchasing behavior

• Seasonality checks
• Correlation analysis
• Data Cleaning
• Data Visualization

Prediction of Store Sales

• Preprocessing
• Building models with sklearn pipelines
• Choice and use of loss function
• Post prediction analysis
• Model serialization

Serving predictions on a web interface

• Basic webpage in HTML and CSS
• Use of flask library for building server side applications
• Preparing scripts for data preprocessing
• Using serialized model for prediction
• Heroku application deployment

While every member of the group performed customer exploration, I was charged with the design and deployment of the web application to Heroku platform.

link to the application.

Project 4 : Ad Campaign Performance with A/B Hypothesis Testing

The project aimed at determininig whether an ad campaign successfully raised awareness. Involves:

• Sample plot derived in the project:

Project 5 : User Analytics in the Telecommunication Industry

Involves:

• Analysing data from a Telecomunication company to drive an investments decision
• Sample plot derived in the project:

Project 6 : African Social Media Influencers Analysis

Involves:

• Web scrapping of influencers from websites
• Searching and downloading Twitter data for the influencers using Twitter API
• Performing sentimental analysis to understand the influence score of each influencer to drive a marketing decision
• A sample plot derived in the project:

Project 7 : Health Data Analysis

Involves:

• Getting data from WHO servers
• Analysing the death rate and the probability of dying between countries using collected data
• Analysing the probability of dying for age 30-70 with NCD
• Analysis of the cases, deaths and recoveries of COVID19 between countries with data from John Hopkins Server
• Sample plot derived in the project: