This was an early version of the PhD work I was focused on during 2024. Since this time, several changes have been made and will eventually be released into an open source project that will argubly, provide more value and greater novelty in terms of value and usage for those wishing to conduct explainable analyis, for now, feel free to browse and look over this early prototype.
VISION is a SHAP-powered local instance investigation tool for time series predictions. The value in this tool is that SHAP investigates global, and LIME investigates local. This tool combines LIME-esque explanations into an interactive timeline for sythentically generated data in a random forest time series forecast. VISION is a powerful tool that improves upon the existing LIME XAI library, whereas previously to investigate several specifc instances of a prediction, you would need a never-ending page of LIME output, VISION allows for an interactive view of local explanations in a user-friendly interface.
https://vision-tool-demo.streamlit.app/
