Model imprv

[jessicarose00]

This branch explores approaches to improve classification accuracy. The approaches include:

• Multi-class classification approach (moving from binary plantation/non-plantation to monoculture/AF/non-plantation)
• Incorporation of texture features
• Deeper dive into feature selection and hyperparameter tuning
• Review of random forest regression performance
• Updates to training data labels

Additional enhancements/new features include:

• exploratory data analysis notebook w/ t-SNE exercise
• restructured src folder to support imports and validation checks
• conclusions/write ups moved to separate notebook
• enhanced visualizations
• steps to improve performance and memory optimization in training and deployment pipeline