IMDb top 1000: from web scraping to metadata-based movie recommender (not personalised)!

I web scraped top 1000 movies on IMDb website, cleaned the gathered data, conducted an exploratory data analysis (EDA), and created a meta-based movie recommender using cosine similarity.

Personalised music artist recommender using SVD (Singular Value Decomposition)

In this project I created a personalised recommendation system. As opposed to the previous project, in this implementation the recommendations take into account the user's personal taste and bias. I did so using Singular Value Decomposition. The final model had an RMSE of 0.01.

Predicting Laptop Prices with Random Forest and Neural Networks!

Before building and testing random forest and neural networks models, I conducted EDA to understand the data at hand better. Random Forest model outperformed my shallow neural network with an R-squared value of 84%.

Classifying chest cancer with 85% accuracy

I built a convolutional neural network with EfficientNetB0 as a base to classify lung images as cancerous, non-cancerous, having a specific type of cancer. Having trained three models, I compared their results and chose the best one to make further improvements on. The final model had 85% accuracy.

News Categorisation

Predicting news category (out of 41 categories) based on their short description (av. length ~11 words) with Naive Bayes and Deep Learning models. The best performing model achieved ~46% accuracy on unseen data with transfer learning.