Data Science Portfolio

Author: Mehdi Outini mehdi.outini@gmail.com

This portfolio is made of several projects showing the work I done in order to develop my data science/machine learning skills.

Table of Contents

• Predicting Housing Prices
• Flowers Recognition
• Bike Sharing Demand

Projects

---

Predicting Housing Prices

Github | nbviewer | kaggle

• Participation of the kaggle competition House Prices: Advanced Regression Techniques
• With 79 explanatory variables describing several aspects of residential homes, we use advanced regression techniques like ElasticNet, random forest, gradient boosting but also multi-layer perceptron to predict the final house price of each home. I used a weighted average of Lasso, ElasticNet, Ridge, XGBoost and Gradient Boosting Regressor to get the best predictor
• The focus of this project was mostly on feature engineering and machine learning models (hyperparameters tuning)

---

Flowers Recognition

Github | nbviewer | kaggle

• Project for the course Deep Learning
• Participation on the dataset Flowers Recognition
• The goal was to classify flower images among 5 species. We use transfer learning with the MobileNetV2 model and demonstrates the efficiency of fine tuning.
• The focus of this project was mostly on computer vision problematics related to classification (deep learning, transfer learning)

---

Bike Sharing Demand

Github | nbviewer

• Participation on the competition Bike Sharing Demand
• The focus of this project was mostly on data vizualisationn and machine learning applied to time series in order to predict the hourly count of bike rentals. The catalog was spanning two years of data.