How 2233-prajyesh/lead-conversion-predection works
Main Components
• Data Input Sources: Represent the three sample data files (SampleData.csv, SampleData.xlsx, synthetic_lead_data.csv). These files are where raw lead data comes from.
• Data Ingestion & Preprocessing Module: This component (implemented in leadconver.py) is responsible for reading the data, cleaning it (handling missing values, duplicates, normalizing column names), and preparing it for analysis.
• Exploratory Data Analysis (EDA) Module: This part handles visualization tasks such as generating histograms, correlation heatmaps, boxplots, and conversion rate graphics. Indicate that Seaborn & Matplotlib are the libraries powering these visualizations.
• Rule-Based Prediction Engine: This module encapsulates the logic that predicts lead conversion. According to the README and model logic, it checks conditions (lead score > 50, email opened = Yes, total time spent > 100 seconds) to produce predictions.
• Performance Evaluation Module: This component computes accuracy by comparing model predictions with the actual conversion data. It outputs performance metrics and can present sample predictions in a tabular format.
Establish Relationships and Data Flows
• From Data Ingestion, create two major output flows:
• One goes into the EDA Module (for visualization and insights gathering).
• The other flows into the Rule-Based Prediction Engine.
Important Architectural Patterns and Technologies
• That the application uses a sequential data processing pipeline where each module feeds into the next.
• That the design uses a modular approach by separating various concerns (ingestion, cleaning, EDA, prediction, evaluation) even though it is implemented in a single Python file. This separation can help guide future refactoring into separate modules if needed.
• Specific libraries (Pandas, Seaborn, Matplotlib) next to the appropriate modules, which helps recognize the technology stack.