Copper Transaction Intelligence Platform utilizing Machine Learning with Python

Project Synopsis & Overview

The Industrial Copper Modeling project is a machine learning-based predictive analytics system designed to improve decision-making within the copper industry by forecasting transaction outcomes and selling prices. The system addresses challenges such as data inconsistencies, missing values, skewed distributions, outliers, and imbalanced datasets commonly found in industrial copper transaction records. The proposed solution utilizes a dataset of daily copper offers containing customer information, product specifications, quantities, pricing details, transaction dates, and transaction status. Data preprocessing includes validation, missing value imputation, feature engineering, logarithmic transformations, normalization, date correction, and outlier handling using the Interquartile Range (IQR) method. The system employs a Random Forest Classifier for predicting transaction status (won/lost) and a Random Forest Regressor for predicting selling prices. Hyperparameter optimization is performed using Grid Search Cross Validation (GridSearchCV) to improve model performance. The major modules include Raw Copper Data, Data Ingestion, Data Preprocessing, Model Training, and Model Registry. The workflow involves collecting raw copper transaction data, preprocessing it, training predictive models, validating performance, and generating predictions through a user-friendly web interface. The expected output is accurate transaction outcome prediction and selling price estimation, enabling enhanced sales strategies, pricing optimization, and business intelligence in the copper industry.

Complete Technical Specifications