Author: Luisa Rosa
Capstone Project

This project analyzes user behavior in an e-commerce platform based on session events (e.g., view, cart, purchase). It extracts meaningful patterns, performs behavioral segmentation, and visualizes trends to uncover insights for user behavior modeling.

The project should be run in five key stages, in this order:

Script: cleaning.py
Cleans the raw data file (2019-Oct.csv) and splits it into smaller chunks with enhanced features:

• Session-level datetime parsing
• price_chunk labels (Low/Medium/High)
• Time-of-day classification (Morning/Afternoon/Night)
• Category simplification via main_category

Output:
Cleaned files saved in ./split_data/ directory.

Script: analysis.py

• Builds user sessions from the cleaned chunks
• Extracts interaction patterns using full session matching and PrefixSpan
• Calculates session-level metrics and cart conversion behaviors

Output:

• deep_dive_patterns.csv
• cart_behavior_metrics.csv
  (saved in ./analysis_results/)

Script: extra.py
Provides a general snapshot of the dataset:

• Event distribution
• Unique users/sessions/brands
• Mapping between brands and categories

Output:

• brand_maincategory_map.csv

Script: plotting.py
Generates visualizations from the pattern summary files:

• Average price per subgroup
• Occurrence heatmaps
• Time-of-day pattern trends

Output:
Multiple .png plots saved in ./analysis_results/

Script: session_analysis.py
Generates summarized session stats + common event patterns.

Output:

• session_analysis_with_patterns.csv

Script: understand_session.py
Analyzes:

• Top session sequences
• Session durations and product diversity
• Browsing-only vs. interactive behavior

Output:
Multiple CSV summaries including:

• pattern_counts.csv
• interaction_sessions.csv
• pattern_stats_top10.csv
  ...and more

Script: understand_patterns.py
Combines and summarizes top patterns per subgroup (price, category, time-of-day).

Output:

• summary_price_chunk.csv
• summary_main_category.csv
• summary_time_of_day.csv

• Python 3.7+
• pandas
• numpy
• matplotlib
• seaborn
• prefixspan

Install dependencies:

pip install pandas numpy matplotlib seaborn prefixspan

• Ensure the raw file 2019-Oct.csv is available in the root directory.
• The cleaning script is capped to process 150 chunks (~7.5 million rows) for speed.
• Modify thresholds or categories directly in the scripts if adapting for other datasets.