Big Data Analytics

Abstract: This research discusses how I implemented a predictive analytics framework using BigQuery Machine Learning (BQML) to address inventory management challenges at Safe Medicare. The primary goal was to optimize medicine stock availability and enhance customer satisfaction through a data-driven approach.

1. Problem Background

Safe Medicare faced significant challenges due to the imbalance between stock availability and patient demand. Fragmented systems led to inefficiencies, such as flu medicine shortages during the rainy season or the overstocking of less popular medications. This not only created a financial burden for the company but also threatened the reputation of its healthcare services.

2. Methodology

In this study, I conducted a comprehensive data processing workflow, which included:

• Data Preparation: Data cleaning (handling null values) and the removal of biased data.
• Feature Engineering: Incorporating features such as month_number to capture seasonal patterns and categorizing waiting times into specific levels (Fast, Normal, Slow).
• Tools: Utilizing BigQuery Machine Learning technology to process large volumes of data in real-time without moving data out of the warehouse.

3. Model and Analysis

• Linear Regression (Demand Forecasting): I utilized this model to predict future drug sales volumes. The model accounts for seasonal trends, promotions, and pricing variables.
• Logistic Regression (Satisfaction Classification): This model was used to predict whether a customer would be satisfied or not, based on variables such as waiting time and medicine availability.

4. Evaluation Results

The model evaluation results demonstrated extraordinary performance:

• The demand prediction model achieved an R² value of 0.997 (near perfect) with a very low Mean Absolute Error (MAE).
• The customer satisfaction classification model reached 100% accuracy (1.0), proving that waiting times exceeding 60 minutes are a critical factor in customer dissatisfaction.

5. Implementation and Conclusion

The implementation of this solution successfully transformed the operational workflow from reactive to proactive. I successfully helped reduce manual staff workload from 15 hours to just 4 hours per week and decreased the risk of expired medicine waste by 65.4%. This system ensures consistent medicine availability for patients.