Boubacar TRAORE

- Adaptation of the Expectation Maximization algorithm to the case of missing data - Implementing the new algorithm in R and C++ for fast computing

Geospatial Analysis: Construction of a quality score on the polygonisation of "Pages Jaunes" locations. - Processing of polygon data - Creation of relevant variables (area, perimeter, shape, etc.) - Construction of matching metrics (Intersection/Union, Rotation Distances, Polys Metric) - Aggregation of metrics for predictive purposes (product, empirical mean, etc.) - Consideration of imbalanced data - Modelling (mainly Logistic Regression and Random Forest Classifier)

Topic: Improvement of anomaly detection algorithms using distributed computing techniques - Introduction to C# and Matlab - Discovering Google Cloud Platform and Spark for Matlab - Adding an automation layer to the anomaly detection algorithm (ML models)

The aim was to carefully analyse the results of the 2017 Survey on Professional Equality and Diversity in the French Ministry of Culture in order to derive relevant conclusions.

Topic : Web Scraping and Data Mining for modelling used car prices in Tunisia To establish a good database, my first work was the automatic extraction of prices and characteristics of used cars on the most popular advertisement site in Tunisia ("Tayara") thanks to the "rvest" library of R (programming language). Thanks to this database, I have set up a "R shiny" dashboard allowing end-users to estimate the price of their used car according to the characteristics provided (brand, model, kilometerage, energy, horsepower, etc...). To achieve this result, I created a text processing pipeline (using Text Mining tools with R) on the comments found in the advertisements to extract some useful information.

This internship was done in the Central Department of National Accounts of INS (National Institute of Statistics). The aim of this internship was double : - Achieve a first immersion in the professional world and also to - Work on a specific issue within this department. My work was focused on the different methods of computing the Fixed Capital Consumption (CCF) which is a key indicator of the depreciation of the value of the fixed's stock assets during a year. This first experience was really interesting for me as I was able to assimilate several notions in economics and understand the practical realities of the professional world.