Topsis-Krit-102303213

TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) command-line package by Krit Mukul (Roll 102303213).

Repository: https://github.com/KritMukul/TOPSIS

Images

Weighted Matrix

Code Example

Installation

From PyPI (after publishing)

pip install Topsis-Krit-102303213

Local development

git clone https://github.com/KritMukul/TOPSIS.git
cd TOPSIS
pip install -e .

Usage

Command-line interface

python -m topsis <InputDataFile> <Weights> <Impacts> <OutputResultFile>

Or after installation:

topsis-cli data.csv "1,1,1,2" "+,+,-,+" output.csv

Parameters

• InputDataFile: CSV file with first column as identifier, remaining columns as numeric criteria
• Weights: Comma-separated numeric values (e.g., 1,1,1,2)
• Impacts: Comma-separated + (beneficial) or - (cost) for each criterion (e.g., +,+,-,+)
• OutputResultFile: Path for result CSV with added Topsis Score and Rank columns

Example

topsis-cli input.csv "1,1,1,1" "+,+,-,+" result.csv

Methodology

TOPSIS Algorithm Steps

This project implements the TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) method for multi-criteria decision making. The methodology is as follows:

1. Normalization: The decision matrix is normalized using vector normalization so that all criteria are dimensionless and comparable.
   • Formula: $r_{ij} = \frac{x_{ij}}{\sqrt{\sum x_{ij}^2}}$
2. Weighting: Each normalized criterion is multiplied by its assigned weight to reflect its importance.
   • Formula: $v_{ij} = w_j \times r_{ij}$
3. Ideal Best & Worst: For each criterion, determine the ideal best (maximum for beneficial, minimum for cost) and ideal worst (opposite).
   • Beneficial (+): best = max, worst = min
   • Cost (-): best = min, worst = max
4. Separation Measures: Calculate the Euclidean distance of each alternative from the ideal best and worst.
   • $S_i^+ = \sqrt{\sum (v_{ij} - v_j^+)^2}$
   • $S_i^- = \sqrt{\sum (v_{ij} - v_j^-)^2}$
5. TOPSIS Score: Compute the relative closeness to the ideal solution.
   • $C_i = \frac{S_i^-}{S_i^+ + S_i^-}$
6. Ranking: Alternatives are ranked based on their TOPSIS scores (higher is better).

All these steps are implemented efficiently using NumPy and Pandas in topsis/core.py.

Example Result Table

After running TOPSIS, the output CSV will look like this (see output.csv):

Fund Name	P1	P2	P3	P4	P5	Topsis Score	Rank
M1	0.67	0.45	6.5	42.6	12.56	0.215	8
M2	0.60	0.36	3.6	53.3	14.47	0.468	4
M3	0.82	0.67	3.8	63.1	17.10	0.861	1
M4	0.60	0.36	3.5	69.2	18.42	0.589	3
M5	0.76	0.58	4.8	43.0	12.29	0.451	5
M6	0.69	0.48	6.6	48.7	14.12	0.327	7
M7	0.79	0.62	4.8	59.2	16.35	0.695	2
M8	0.84	0.71	6.5	34.5	10.64	0.414	6

Where:

• Topsis Score: The relative closeness to the ideal solution (0 to 1)
• Rank: 1 is best

Example Input Table

The input data (see test-data.csv) should look like:

Fund Name	P1	P2	P3	P4	P5
M1	0.83	0.69	5.8	41.4	12.18
M2	0.83	0.69	5.8	63.0	17.58
M3	0.84	0.71	3.7	32.5	9.44
M4	0.72	0.52	6.0	70.0	19.31
M5	0.73	0.53	4.7	65.5	17.87
M6	0.75	0.56	6.5	63.6	17.85
M7	0.80	0.64	3.6	45.6	12.66
M8	0.79	0.62	3.3	60.8	16.38

Result Graph

You can visualize the results using a bar plot of TOPSIS scores or ranks. Here is an example using matplotlib:

import pandas as pd
import matplotlib.pyplot as plt

df = pd.read_csv('output.csv')
plt.figure(figsize=(8,4))
plt.bar(df['Fund Name'], df['Topsis Score'], color='skyblue')
plt.xlabel('Alternative')
plt.ylabel('Topsis Score')
plt.title('TOPSIS Scores by Alternative')
plt.show()

This will produce a graph showing the relative performance of each alternative. You can also plot the ranks for a clearer comparison.

Input Format

CSV file with:

• Header row
• First column: Alternative names/IDs
• Remaining columns: Numeric criteria values (minimum 2 criteria required)

Output Format

Input data with two additional columns:

• Topsis Score: Closeness coefficient (0-1)
• Rank: Rank based on score (1 = best)

Publishing to PyPI

Build

python -m build

Upload

python -m twine upload dist/*

Test installation

pip install Topsis-Krit-102303737
topsis-cli --help

License

MIT License - see LICENSE file

Author

Krit Mukul - Roll No. 102303213