Random Forest was analyzed through Cross Validation. Too little improvement was significant; therefore, different approaches were used to analyze the output and data. For example, neural network and K-means Clustering.
Approaches accomplished:
1. K-Means Clustering
[x] Target on "Max_tib_lat_contact_pressure" and "Max_tib_med_contact_pressure," no frame number at this time.
[x] Cluster of 3 centroids. Based on minimal, Medium, and High stress from Max pressure data.
[x] Both three levels are heavily weighted and influenced by the dataset of bones with normal pressures.
[x] Analysis result was very similar to the outcomes of random forest. Both predictions indicate the outcomes not based on the frame number were still heavily influenced.
[x] It's safe to conclude, at this point, that the data itself is not well generated, and most likely, this doesn't require modeling because stress peak can work and be done by data science technique.
2. Neural Network
[x] After K-Means Clustering, I was hoping to generate another model to support the outcomes of random forest. Therefore NN was raised once again, given that it has a similar classification process to tress.
[x] Huge Iteration was running on the cloud with large batch size.
[x] Break the dataset with no frame number and randomize the plot to generate a better result.
harrybsu123
commented
1 year ago
Random Forest was analyzed through Cross Validation. Too little improvement was significant; therefore, different approaches were used to analyze the output and data. For example, neural network and K-means Clustering.
Approaches accomplished:
1. K-Means Clustering
2. Neural Network