erdos2n/pcavstsne

I compared PCA and TSNE in a sequential neural net to see which dimensionality reduction algorithm would produce better results.

PCAvsTSNE

This will compare PCA and TSNE in a sequential neural net and evaluate the score for each

reduction algorithm.

---

Without dimensionality reduction:

• evaluation score = 0.66
• PCA score = 0.77
• TSNE score = 0.70

---

Looking at the table below there is already a baseline of predicting 65% zeros, so the real

question is whether or not this can predict a one (diabetic). Further analysis needs to be done

to answer this question and I plan on running a random forest classifier to see if it produces better

results.

	1	0	Total
y	268	500	768
%	35%	65%	100%

---

There will also be a plot.ly diagram coming showing the labeled data in 3d space.

Any and all advice/insights are welcome!

Thanks!