Skip to content

CogitoNTNU/CogitoXInfor-h24

Repository files navigation

Cogito x Infor Project

Cogito Image

Overview

This project is a collaboration between Cogito, a student organization, and Infor, the third-largest company globally for enterprise systems. The goal was to develop a dynamic pricing model using data science techniques and to integrate insights into EA Smith's pricing strategy for improved profitability.

Prerequisites

  • Ensure that Git is installed on your machine. Download Git
  • Python 3.7 or higher is required to run the Jupyter notebooks. Download Python

Quickstart Guide for Jupyter Notebooks

This guide provides an overview of the uploaded Jupyter notebooks and how to get started with them.


Available Notebooks

1. FastMarketsAPI/APICall.ipynb

  • Purpose: Demonstrates how to interact with an API to fetch data.
  • Features:
    • API key authentication.
    • Sending requests and parsing JSON responses.
    • Error handling for failed API calls.

Quickstart

Contact EA. Smith, FastMarkets or Thomas for API KEY

  1. Ensure Python and required libraries are installed:
    pip install requests

2. PriceRecommender/FinalNotebookCogito.ipynb

The FinalNotebookCogito.ipynb notebook is a key component of the Price Recommender Project. It focuses on creating a dynamic pricing recommendation model through advanced data processing, clustering, and evaluation techniques.


Key Features

1. Data Preprocessing

  • Input Dataset:
    • Source: ECOMMRecords 2020 (or similar structured data).
    • Content: Product and customer information for pricing insights.
  • Cleaning Steps:
    • Handle missing data.
    • Normalize numerical features.
    • Encode categorical variables.

2. Feature Engineering

  • Extracted key features from the dataset for clustering.
  • Feature selection includes:
    • Product category.
    • Customer segment.
    • Historical purchase behavior.

3. Clustering Analysis

  • K-Means Clustering:
    • Group products and customers into clusters based on similarities.
  • Evaluation:
    • Silhouette score to determine the optimal number of clusters.
    • Cluster visualization to ensure meaningful segmentation.

4. Price Recommendation Model

  • Uses the clusters to recommend optimal prices.
  • Customizes margins for each product-customer combination.

5. Visualizations

  • Graphical representations of clusters.
  • Distribution plots for customer and product clusters.
  • Insights into recommended pricing adjustments.

How to Use

Prerequisites

  • Dependencies: Install the required Python libraries using:
    pip install pandas numpy scikit-learn matplotlib seaborn
    

Problem

EA Smith faced challenges with:

  • Unclear pricing strategies
  • Limited market insights

This resulted in poor pricing decisions and reduced profit margins.


Solution

The team developed a pricing recommendation model using unsupervised learning to identify natural clusters in product and customer data. Key benefits of this approach include:

  • No need for historical data
  • Tailored margin recommendations for product-customer combinations
  • High interpretability

Methodology

Data Collection

  • Dataset: ECOMM Records 2020 from Kaggle
  • Additional scraping via APIs to gather competitor pricing.

Modeling Approach

  1. Feature Engineering:
    • Customer and product attributes were processed for clustering.
  2. K-Means Clustering:
    • Analyzed clusters and optimized the model using silhouette scores.
  3. Pricing Recommendations:
    • Margins were adjusted dynamically per cluster.

Competitor Analysis

  • Developed an API scraper to:
    • Compare EA Smith's prices with competitors.

Currency Conversion

  • Integrated an API to convert prices from EUR to NOK for localized analysis.

Key Features

  • Dynamic Pricing Model:
    • Custom margin recommendations using unsupervised learning.
  • API Integration:
    • Fetch competitor prices dynamically.

Results

  • Improved Pricing Strategies:
    • Tailored price recommendations by cluster.
  • Market Insights:
    • Clear competitor comparison through API scraping.

Future Plans

  • Continued collaboration with Infor for extended features.
  • Enhanced dynamic pricing models for broader adoption.

Team

  • Project lead: Thomas
  • Members/Developers: Herman, Baris, Nikolai, Gard
Thomas Sørensen
Thomas Sørensen
Baris Batur
Baris Batur
Gard
Gard
Herman
Herman
Nikolai Helgås Helleseth
Nikolai Helgås Helleseth

Acknowledgments

In collaboration with Infor and EA. Smith

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published