M.S. AAI Capstone Chronicles 2024

Figure 3.4: The scatter plot shows the relationship between actual and predicted carbohydrate values, with the red dashed line (y=x) representing ideal predictions. Most points are near the ideal line, but there is some spread, especially at higher true values, indicating that the model’s predictions deviate more as carbohydrate levels increase. The box plot of prediction errors shows that the errors are generally centered around zero, indicating minimal systematic bias. However, there are a few outliers on both the positive and negative sides, highlighting instances where the model struggled to predict accurately. The interquartile range (IQR) shows that the middle 50% of the errors are relatively small, but the overall error spread suggests room for improvement in reducing variance and addressing edge cases. This can be achieved by adding more data samples at the low and high ends of the carb food dish distribution.

23

257