Exploring data synthesis using diffusion modeling to create a more representative dataset, which is being used to train an RL controller for microgrids.

Building a system for ARC Hatchery to count larvae and identify them as healthy/unhealthy.

Creating an ML model suite, which will be used to optimize collections strategies.

Debt Repayment Prediction Model for AG Adjustments and HoneywellObjective:Predict how likely Honeywell's debtors are to repay their debts.Process:Data Preparation: Loaded and cleaned the dataset for better accuracy. Created new features that would help in making precise predictions.Data Split: Divided the data into parts for training and validating the model's performance.Addressing Imbalances: Implemented the SMOTEENN method to ensure that predictions are balanced and accurate.Model Selection: Opted for the XGBoost classifier, known for its high accuracy in similar prediction tasks.Feature Selection: Pinpointed which data points or features most impact the predictions.Model Tuning: Adjusted settings for the best results, with the final model achieving an impressive 94.3% accuracy.Outcome:This model is planned to be a crucial component of a new product suite by AG Adjustments, set for a year-end launch. It will aid Honeywell in more informed financial decision-making. Due to the proprietary nature of the model, technical details must remain general.