This project focuses on the classification of banknotes using various supervised machine learning models. The primary objective is to develop a robust system that can accurately distinguish between genuine and counterfeit banknotes based on a set of features. The dataset used for this project undergoes thorough exploratory data analysis (EDA), visualization, and data cleaning processes to ensure the quality of the input data.
Target Audience: This project is beneficial for banks, financial institutions, and authorities involved in currency circulation and fraud detection. The classification model can assist in automating the detection of counterfeit banknotes, enhancing the efficiency and accuracy of the overall verification process
- Machine learning
- Data CLeaning and Processing
- Model Building
- Feauture Engeeneering
- Data Visualization
- Model Evaluation
- Exploratory Data Analysis (EDA):
In-depth analysis of the dataset to understand its structure and characteristics.
Identification of key features and their distributions. Handling missing values and outliers to ensure data quality.
- Data Cleaning:
Preprocessing steps to clean and prepare the dataset for model training. Addressing any inconsistencies, errors, or anomalies in the data.
- Visualization:
Visual representation of data trends and patterns to gain insights. Correlation analysis to identify relationships between features.
- Modeling:
In banknote classification, diverse machine learning models exhibited remarkable accuracies.
Logistic Regression demonstrated strong performance at 98.36%,
Support Vector Machine with a linear kernel achieved 98.54%. the Support Vector Machine with an Rfc (Radial basis function) kernel achieved perfect accuracy at 100%.
The Random Forest model showcased exceptional accuracy, reaching 99. 27%
the K-Nearest Neighbors (KNN) model excelled with an outstanding accuracy of 99.82%.
These results underscore the effectiveness of different algorithms in accurately discerning between genuine and counterfeit banknotes.
- Evaluation:
Comparing the performance of different models and creating a sns Confusion Matrix