This week was about momentum and clarity. We aligned on the big picture, turning feedback into a sharper plan, and committed to a supervised learning direction for replacement prediction, with a stronger grounding in industry practices. That focus now guides our feature choices, validation approach, and prep for SLDR.

We sharpened our model direction this week: after PID feedback, we explored “mean lifetime” and industry practices (LM-80/TM-21) and chose a supervised approach for predicting replacement timing, using warranty and reference life as priors. We also clarified how vendors estimate expected life (LM-80 → TM-21), which guides our features and validation. On the UI side, Our team refined the flow (moving login confirmation), added colored bulb status to floor plans, linked product bands to purchasing, and surfaced brand/type with warranty life. Meanwhile, we compared backend frameworks to integrate models, database, and dashboard cleanly, sketching service boundaries and communication.

We finalized the SLDR structure and began drafting. The narrative follows Problem → Approach → Value, with concise sections on user needs and data sources (tested vs. reported specs). We’re assembling demo visuals: a clean system diagram (models ↔ database ↔ dashboard) and a floor-plan view where fixtures display predicted lifetime.

Moving Forward!!

we’re rolling updates into the Streamlit prototype based on recent feedback, for a smoother, and interactive UI. In parallel, we’ll continue model development with the new insights we gathered and visualize the model’s internal behavior and feature effects using Python-based explainability and plotting tools.

Reflection

This week we learned the power of focus and storytelling. Listening to feedback helped us see the big picture and choose a clearer path. We also learned how important it is to explain ideas simply, what matters to building managers, not just to us, and to keep the experience intuitive on the page.