Dataset Overview

Dataset Summary

The dataset analyzes historical POS data after automation, with an emphasis on PIX payment transactions. Heatmaps visualize transaction values across time frames and metrics, revealing transaction patterns.

Objectives

• Evaluate transaction behavior in a post‑automation POS environment.
• Detect anomalies or patterns in transaction values across different time windows.
• Offer visual insights through heatmaps to help stakeholders quickly grasp trends and focal points.

Technology Stack

• Python 3.9+: Core language for analysis and visualization.
• Pandas: Data manipulation and analysis.
• Matplotlib & Seaborn: Generation of heatmaps and other visualizations.
• NumPy: Numerical computation.
• Flask: Backend framework for serving data and APIs.
• PostgreSQL: Database storing transaction records.
• Docker: Containerization for reproducible environments.
• JWT Authentication: Secures API endpoints that provide transaction data.
• Qdrant: Vector database for optimized query handling.
• YAML: Configuration management for servers and endpoints.

Heatmaps

Heatmap 1: Maximum Adjusted Transaction Value

The red areas indicate transaction values markedly higher than the average, suggesting peak usage periods or potential anomalies.

Heatmap 2: Minimum Adjusted Transaction Value

This heatmap highlights periods where transaction amounts are significantly lower, possibly indicating low‑traffic intervals or operational inefficiencies.

Heatmap 3: Average Transaction Value

Orange or red zones denote higher average values; lighter colors represent lower averages.

Heatmap 4: Maximum Transaction Value

Red regions show periods when the system processed large payments.

Heatmap 5: Minimum Transaction Value

This map provides insight into the lowest transaction values, useful for pricing analysis and understanding customer behavior.

Data Processing Workflow

1. Data Ingestion: Extract relevant PIX transaction records from the PostgreSQL database.
2. Pre‑processing: Clean and format data using Pandas, handling missing or inconsistent entries and engineering date‑time features.
3. Metric Computation:
   • Maximum Adjusted Transaction: Normalize transaction values considering inflation or operational changes.
   • Minimum Adjusted Transaction: Similar to the maximum but focusing on the lower bound.
   • Average Transaction: Compute mean values per time window.
   • Maximum & Minimum Transaction: Capture extreme transaction values.
4. Visualization: Generate heatmaps with Matplotlib and Seaborn.