Week 12 and 13 updates

 

Blog Update 3

Week 12 and 13 20/05/25 - 03/06/25

Inference Pipeline Development and Final Presentation Preparation

This week marked the transition from model experimentation to a fully functional inference pipeline, laying the groundwork for real-time glucose prediction. Alongside technical implementation, we also finalized the presentation materials and prepared a demo package for showcasing the end-to-end system.

Inference Pipeline Implementation

To simulate real-world prediction scenarios, a test sample was extracted and pushed to the end of the dataset. This allowed us to mimic unseen data inference while preserving a ground truth value for validation. The complete dataset, including this test sample, was then passed through the full preprocessing workflow, which included normalization and dimensionality reduction.

Curvilinear Component Analysis (CCA) was used to transform the feature space, reducing it to 45 dimensions. This technique, already proven effective in prior phases, ensured that the low-dimensional representation maintained essential structure and relationships within the data.

Once the transformation was complete, the corresponding glucose value for each sample was appended to the reduced dataset. This enabled the pairing of transformed feature vectors with known glucose levels for both training and testing.

Model Testing and Real-Time Estimation

With the inference pipeline in place, the test sample was used to evaluate the performance of our trained linear regression model. The model generated a glucose level estimate, which was compared to the known value. The resulting estimation error provided a realistic measure of the model’s effectiveness in a single-sample prediction context.

To improve generalization and reliability, we retrained the regression model using the complete CCA-transformed dataset. This newly trained inference model was then used to predict the glucose level of the same test sample. The new estimate was again compared with the true value, and the absolute difference was recorded as the inference error.

Model Performance Evaluation

In addition to single-sample testing, the updated model was evaluated across the entire dataset to better understand its overall accuracy. Predictions were made for all available samples, and the errors between predicted and actual glucose values were computed.

Two primary metrics were used for performance evaluation:

• Mean Absolute Percentage Error (MAPE): Assesses the average size of errors as a percentage of the predicted values.

• Root Mean Square Error (RMSE): Provides a sense of overall deviation, penalizing larger errors more heavily.

Together, these metrics offered a comprehensive picture of how well the model performs under both localized and global prediction conditions. They also served as a benchmark to compare against future model refinements.

Presentation and Demo Preparation

With the technical work nearing completion, we shifted attention to presentation and documentation. A PowerPoint presentation was created to effectively communicate the project scope, methodology, technical challenges, and key findings. The slides walk through each major phase—from data acquisition and preprocessing to dimensionality reduction and machine learning implementation.

In parallel, all final demo files were compiled. This included the cleaned datasets, trained models, output visualizations, and performance metrics, ensuring a smooth and complete demonstration experience. The goal was to showcase both the predictive capability and practical potential of our system in a clear and impactful way.