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.
- Ensure that Git is installed on your machine. Download Git
- Python 3.7 or higher is required to run the Jupyter notebooks. Download Python
This guide provides an overview of the uploaded Jupyter notebooks and how to get started with them.
- 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.
Contact EA. Smith, FastMarkets or Thomas for API KEY
- Ensure Python and required libraries are installed:
pip install requests
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.
- Input Dataset:
- Source:
ECOMMRecords 2020(or similar structured data). - Content: Product and customer information for pricing insights.
- Source:
- Cleaning Steps:
- Handle missing data.
- Normalize numerical features.
- Encode categorical variables.
- Extracted key features from the dataset for clustering.
- Feature selection includes:
- Product category.
- Customer segment.
- Historical purchase behavior.
- 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.
- Uses the clusters to recommend optimal prices.
- Customizes margins for each product-customer combination.
- Graphical representations of clusters.
- Distribution plots for customer and product clusters.
- Insights into recommended pricing adjustments.
- Dependencies:
Install the required Python libraries using:
pip install pandas numpy scikit-learn matplotlib seaborn
EA Smith faced challenges with:
- Unclear pricing strategies
- Limited market insights
This resulted in poor pricing decisions and reduced profit margins.
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
- Dataset: ECOMM Records 2020 from Kaggle
- Additional scraping via APIs to gather competitor pricing.
- Feature Engineering:
- Customer and product attributes were processed for clustering.
- K-Means Clustering:
- Analyzed clusters and optimized the model using silhouette scores.
- Pricing Recommendations:
- Margins were adjusted dynamically per cluster.
- Developed an API scraper to:
- Compare EA Smith's prices with competitors.
- Integrated an API to convert prices from EUR to NOK for localized analysis.
- Dynamic Pricing Model:
- Custom margin recommendations using unsupervised learning.
- API Integration:
- Fetch competitor prices dynamically.
- Improved Pricing Strategies:
- Tailored price recommendations by cluster.
- Market Insights:
- Clear competitor comparison through API scraping.
- Continued collaboration with Infor for extended features.
- Enhanced dynamic pricing models for broader adoption.
- Project lead: Thomas
- Members/Developers: Herman, Baris, Nikolai, Gard
 Thomas Sørensen |
 Baris Batur |
 Gard |
 Herman |
 Nikolai Helgås Helleseth |
In collaboration with Infor and EA. Smith