Portfolio

Data Science and Machine Learning Projects

‘Prepaid Mobile Subscribers Segmentation’ was a comprehensive data-driven project designed to enhance the service offerings of Tunisie Telecom by effectively segmenting their customer base. In the pursuit of providing superior services, Tunisie Telecom continuously accumulates extensive data through their networks and systems, offering detailed insights into customer behavior. The overarching business objective of this project was to tailor prepaid mobile plans to individual customers based on their behavioral patterns.

Our project harnessed a representative sample of this data, documenting the activities and preferences of 27,565 anonymous customers during the critical months of December 2016 and January 2017. To achieve customer segmentation, we employed a consumption-based approach, carefully analyzing and categorizing customers into distinct segments. These segments were elucidated through thorough analysis and presented using a suite of visualizations and interactive dashboards, aimed at elucidating the segmentation criteria and facilitating a clear understanding of the data.

Furthermore, as a key component of our project, we engineered a classification model capable of predicting a customer’s segment based on their behavior and usage patterns. This model serves as a valuable asset for Tunisie Telecom, enabling them to optimize service recommendations for individual customers, thereby enhancing the overall customer experience and maximizing operational efficiency.

Tweets Sentiment Analysis

The ‘Tweets Sentiment Analysis’ project centered around the hashtag #StressAwarenessDay is a testament to the power of data-driven insights in understanding and addressing critical societal issues. To initiate this endeavor, we embarked on a data collection journey by scraping Twitter for tweets using the designated hashtag. This initial step allowed us to collect real-time data, capturing a snapshot of conversations, concerns, and support related to stress awareness on the 7th of November 2018.

With our dataset in hand, we crafted a graph that depicted the intricate web of user interactions under the #StressAwarenessDay hashtag. This graph not only showcased the volume of tweets but also highlighted connections among users, painting a picture of the online community’s interaction and engagement in spreading awareness about stress-related issues.

However, the project’s true depth lay in the application of sentiment analysis. Leveraging natural language processing techniques, we delved into the sentiments expressed within the tweets. This phase enabled us to categorize tweets into positive, negative, or neutral sentiments, shedding light on the emotional undercurrents within the stress awareness dialogue. By examining the sentiment patterns, we gleaned valuable insights into the effectiveness of the campaign, identifying areas of positivity, concern, or potential areas for improvement.

Flight Delay Prediction

‘Flight Delay Duration Prediction’ was a challenge part of the AI Hack Tunisia event in 2022, it was made for Tunisair, the flag carrier airline of Tunisia.

Flight delays, which have long vexed both passengers and airlines, are more than mere irritations; they disrupt schedules, decrease operational efficiency, lead to increased capital costs through crew and aircraft reallocation, and incur additional crew expenses. Beyond these immediate consequences, a consistent record of delays can tarnish an airline’s reputation and impact passenger demand.

Our solution, executed with precision, was dedicated to predicting the estimated duration of flight delays on a per-flight basis. This mission has far-reaching implications for the entire air travel ecosystem.

We achieved our goal by harnessing the formidable capabilities of Machine Learning techniques. Our flight delay predictive model excels in accuracy and empowers all stakeholders within the air travel industry. With precise delay predictions in hand, airlines, airports, and travelers can proactively plan and execute effective strategies, mitigating the disruptive effects of delays. This initiative has already yielded significant savings in terms of time, capital allocation, and resource preservation.