A carregar...
Publicação
iFood Catalog Enhancement: Improving Restaurant Catalog Management and User Recommendations at iFood using Machine Learning Techniques
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.35 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
This thesis presents an in-depth exploration and implementation of advanced Data Science methodologies within the operational context of iFood, a prominent Brazilian online food delivery platform. The principal objective of the study was to augment the management and recommendation processes for items within iFood's extensive restaurant catalog, fostering informed and data-driven decision-making.
The focal point of this endeavor was the development and deployment of machine learning models crafted to address a critical challenge, the lack of standardized, centralized information within the platform's catalog due to the diverse and unstructured data input from restaurants. The primary emphasis lay in taxonomy classification, aiming to systematically categorize menu items into a structured framework. Guided by principles of modularity, accuracy, cost-effectiveness, and the MECE (Mutually Exclusive, Collectively Exhaustive) principle, the classification ensured a comprehensive and non-overlapping categorization of items.
Key to the project's success was the handling of data drift, a phenomenon where the performance of machine learning models degrades over time due to changes in input data distributions. The study emphasized the importance of continuous model monitoring and the integration of “Human in the Loop” (HITL) systems for real-time error correction and model validation. This approach ensured that the models remained accurate and relevant in the dynamic environment of the food delivery industry.
The thesis employed a range of techniques including Logistic Regression, XGBoost, FoodBERT (a specialized adaptation of Google's BERT model for the food domain), and TF-IDF vectorization. The project's comprehensive nature extended beyond model accuracy, encapsulating broader business and operational considerations such as cost, interpretability, and response time, reflecting the complex demands of a corporate setting.
Descrição
Internship Report presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Data Science
Palavras-chave
Taxonomy Classification Text Classification Machine Learning Restaurant Catalog Optimization Human In The Loop