Those challenges come from:
I've achieved varisous tasks / used different types of ML models:
- regression / classification
- computer vision
- natural language processing
- recommendation engine
| Title & notebook | Year | Source | Description & Goals |
|---|---|---|---|
| Fashion MNIST | 2019 | Kaggle | Classification task of 28x28 grayscale clothes' images with CNNs & MLPs in Tensorflow |
| Bike Sharing Demand | 2019 | Kaggle | Forecast rentals of a city bikeshare system with datetime, weather, rentals number infos (supervised ML regression) |
| House Prices | 2019 | Kaggle | Use case: predict sales prices (Supervised ML with feature engineering) |
| Dog or Cat | 2019 | ? | Distinguish images of dogs from cats : binary classification / computer vision using CNN with Tensorflow |
| Adult Census Income | 2019 | Kaggle | Predict whether income exceeds $50K/yr based on census defining people profiles (supervised ML binary classification) |
| Credit Default Risk | 2019 | Kaggle | Prediction whether or not an applicant will be able to repay a loan based on previous credits, POS (point of sales), cash loans, repayment history... |
| Hybrid Recommendation Engine | 2019 | Kaggle | Recommandation System with LightFM (collaborative & content-based filtering) trained on the MovieLens 100K dataset |
| Customer Segmentation | 2019 | Kaggle | Segmentation : target customers with whom you can start marketing strategy (unsupervised ML - KMeans Clustering) |
| Fraud Detection | 2019 | Kaggle | Anomaly detection among anonymized credit card transactions labeled as fraudulent or genuine (PCA than oversampling on highly imbalanced classes (SMOTe)) |
| Real Estate Price | 2019 | Kaggle | Prediction of house prices for California districts derived from a census (analysis of geospatial data) |
| Customer Churn | 2019 | Kaggle | Predict behavior to retain your customers before they live |
| Historical consumption regression for electricity supply pricing | 2020 | E.N.S | Predict the electricity consumption of 2 given sites during the next year using time series & deep learning models R.N.N / G.R.U. |
| Solar Power Generation in EU - Part 1 / 2 / 3 | 2020 | Kaggle | Countries segmentation depending on their generation profile, power plant's efficiency prediction accross time with R.N.N & Prophet. |
| Cirta - Particle Type Classification - Part 1 & 2 | 2020 | Zindi | Build a machine learning model to help physicists identify particles after a collision based on images. |
| Handwritten Digit Generation | 2021 | Kaggle | Unsupervised deep learning with G.A.N - training of a Generator and a Discriminator with the famous MNIST dataset |
| Wind Power Generation in EU - Part 1 / 2 | 2021 | Kaggle | Countries segmentation depending on their generation profile, efficiency prediction analysis. |
| Fowl Escapades | 2022 | Zindi | Classification of numerious classes of birds callings. Audio data augmentation, preprocessing, spectrograms & deep learning with CNNs |
| Pima Indians Diabetes - Analysis & Predictions | 2022 | Kaggle | Analysis of patients observations, health measures and how to build a machine learning model to diagnostically predict whether or not a patient has diabetes |
| Treadmills users - Advanced data viz, recommendation) | 2022 | Kaggle | Interactive & dynamic charts with Plotly with advanced stats, recommendation & clustering of mix data types with K-prototypes |
| Prediction of building's greenhouse gas emissions (GHG) - part 1 - EDA & 2 - Modelling | 2023 | Kaggle | Geospatial analysis - regression & explainability |
| H&M Recommendation System for Personalized Fashion - part 1 - EDA & 2 - Modelling | 2023 | Kaggle | Analysis of datasets with RecoSys specific caracteristics, and fine tuning of an hybrid lightfm model |