PythonScikit-learnBERTNetworkXScrapy
Food Delivery Patterns & Smart Recommendation System
Led the visualization & reporting stream of a 5-person team analyzing 2,000 orders to power a hybrid recommendation engine.
📊 Top meal score: 1.00 | 620 association rules
Dataset
2,000 food delivery orders across 487 customers, 20 restaurants, and 1,835 customer reviews — collected via Scrapy & BeautifulSoup.
Methodology
- 1Apriori association rule mining → 620 co-ordered meal combinations.
- 2Co-ordering graph (355 nodes, 5,890 edges) ranked with PageRank and HITS.
- 3BERT sentiment analysis on 1,835 reviews.
- 4Unified score: 70% sentiment × 30% network influence.
Final Outcome
Top-ranked meal achieved a final recommendation score of 1.00. The hybrid system combines what customers say with what they actually order together.
Skills Used
PythonScikit-learnBERTNetworkXScrapyBeautifulSoupNLP