Sami Khenissi
Research scientist @ Meta
Archives
Predicting Estimated Repair Time for Power Outages for LG&E and KU Energy
I designed a Data Science pipeline to predict the ETR (Estimated Repair Time) for different power outages. The model improved the prediction of the company by 50%
Theoretical Modeling of the Iterative Properties of User Discovery in a Collaborative Filtering Recommender System
The closed feedback loop in recommender systems is a common setting that can lead to different types of biases. Several studies have dealt with these biases by designing methods to mitigate their effect on the recommendations. However, most existing studies do not consider the iterative behavior of the system where the closed feedback loop plays…
Modeling and Counteracting Exposure Bias in Recommender Systems
What we discover and see online, and consequently our opinions and decisions, are becoming increasingly affected by automated machine-learned predictions. Similarly, the predictive accuracy of learning machines heavily depends on the feedback data that we provide them. This mutual influence can lead to closed-loop interactions that may cause unknown biases which can be exacerbated after…
An end-to-end intelligent assistance tool for crowdfunding projects
Estimate the probability of success of crowdfunding projects using machine learning classification algorithms and develop a Ruby library to help users boost their campaigns. This work has been done during my Data Science Internship at DigiSponsor in Paris Collect (scrape) crowdfunding project’s data from Kickstarter using python Perform exploratory statistics and analysis on the data…
Explainable Active Learning Strategy For Recommender Systems
Recommender System technologies are witnessing a revolutionary era these days. They are growing more and more important since they help in the discovery and promotion of products, books, news, music, movies, courses, restaurants, etc. One major problem of Recommender Systems is that the most accurate models tend to be black boxes, whose results cannot be…
Predict credit card Frauds using Anomaly Detection techniques (Kaggle competition – Silver Medal)
Participated in an online Kaggle competition with an aim to detect fraudulent credit card transactions. The main focus was to learn and understand how to properly validate a Machine Learning model. To avoid overfitting and to make sure that our Cross-Validation technique is reliable, we explored many validation strategies. We also implemented different ensembling techniques…
Implementing an Online Recommender system based on MovieLens data and ExplainableMatrix Factorization algorithm
Team Leader of the Project (team of 5 students): Track progress and design next steps based on Agile process Using JEE, we implemented a web application that manages accounts and rating systems Using python, we designed a web service to provide explainable recommendation using REST architecture that we hosted on AWS Write Different design documents…