Discrete Structures Semester Project

Price Similarity Network Analysis using Pakistan Bureau of Statistics CPI Data

Overview

This desktop application analyzes multi-year Consumer Price Index data to construct temporal similarity networks between cities based on their price patterns across different product categories.

Features

• Multi-year temporal network analysis
• 7 product categories
• 4 centrality metrics (Degree, Closeness, Betweenness, Eigenvector)
• Multiple weighting schemes (Equal, Entropy-based, Correlation-based)
• Temporal partial order verification
• Interactive visualizations
• Data aggregation across years

Requirements

• Python 3.8 or higher
• See requirements.txt for package dependencies

Installation

Windows

INSTALL.bat

Or manually:

python -m pip install -r requirements.txt

Linux / macOS

pip install -r requirements.txt

Usage

Windows

RUN_APP.bat

Or:

python desktop_app.py

Linux / macOS

python3 desktop_app.py

Data Format

Input CSV file must contain the following columns:

• Year - Numerical year
• Month - Month name
• City - City name
• Product - Product name
• Price - Numerical price value

Default data file: data.csv

Application Structure

Main Components

• desktop_app.py - GUI application
• price_network_analysis.py - Core analysis engine
• data.csv - CPI dataset

Interface Tabs

1. Overview - Dataset statistics with year filtering
2. Network - Interactive graph visualization
3. Centrality - Metric comparison charts
4. Rankings - City influence scores
5. Heatmap - Similarity matrices
6. Temporal - Partial order analysis
7. Credits - Project information

Parameters

• Similarity Threshold (0.5 - 0.95)
• Normalization Method (Z-score, Min-Max)
• Missing Data Strategy (Drop, Mean, Median)
• Weighting Scheme (Equal, Entropy-based, Correlation-based)

Mathematical Foundation

Cosine Similarity

similarity(A, B) = (A · B) / (||A|| × ||B||)

Weighted Centrality Aggregation

S(i,y,c) = w_D × D(i,y,c) + w_C × C(i,y,c) + w_B × B(i,y,c) + w_E × E(i,y,c)

Entropy-Based Weighting

H_k = -Σ(p_ik × log(p_ik))
w_k = H_k / Σ(H_j)

Temporal Relation

G(y,c) ⊆ G(y',c) defines partial order

Data Aggregation

When "All Years" is selected:

• Networks: Combines graphs and averages edge weights
• Centralities: Averages metrics across years
• Rankings: Aggregates and averages influence scores
• Heatmaps: Element-wise average of similarity matrices

Export Features

• Rankings can be exported to CSV
• Visualizations displayed in application

Project Information

Course: Discrete Structures
Institution: FAST National University
Program: BS Computer Science
Semester: Fall 2024

Team Members:

• Hassaan Amin (24I-0880)
• Ammar Bin Omer (24I-0500)
• Haris Zahid Abbasi (24I-0643)

Technical Stack

• CustomTkinter - GUI framework
• NetworkX - Graph analysis
• Pandas - Data processing
• NumPy - Numerical computing
• Matplotlib - Plotting
• Seaborn - Statistical visualization
• SciPy - Scientific computing

License

Academic project for educational purposes.

Data Source

Pakistan Bureau of Statistics (PBS)
Consumer Price Index Publications
https://www.pbs.gov.pk/price/