Capstone Project Final code file is very large, therefore the drive link is provided here: https://colab.research.google.com/drive/13iSDqdbjhq85HudLf-xO08-x7SQJcmxk
Profile report for train dataset is large file , therefore the drive link is provided here: https://drive.google.com/drive/folders/1NZMNBS2fExX3yh1ga_67t3xR_GeiXRzd
Profile report for test dataset is large file , therefore the drive link is provided here: https://drive.google.com/drive/folders/1NZMNBS2fExX3yh1ga_67t3xR_GeiXRzd
Steps
- Choosing Dataset and Theme Selection of the Sign Language MNIST dataset from Kaggle for the project theme.
- Cleaning/Preparing Data Preparing and revising EDA report, and reviewing existing literature on the dataset.
- Initial Problem Analysis Writing a literature review with a focus on the research questions.
- Exploratory Data Analysis (EDA) Describing data, identifying missing values and outliers, analyzing attribute types, conducting descriptive statistics, class distribution analysis, and correlating attributes.
- Feature Selection Using a hybrid and filter approach for selecting the most predictive pixels, with a percentage based cutoff.
- Experimental Design and Cross-Validation Selecting six classification algorithms, including CNN.
- Predictive Modelling Evaluating models based on accuracy, precision, recall, and confusion matrix; assessing models on stability, efficiency and effectiveness.
- Conclusion and Recommendation Concluding the analysis and making recommendations.
- Limitations, Future Direction Discussing the limitations of the current study and suggesting directions for future research.