In two previous articles (1, 2) I've explained how I've implemented the training of Support Vector Machines and Neural Network and shown that I obtain good results on several datasets. Training is great, but if there is no way to use the trained model that's pretty useless. So I implemented right away an export of the prediction as a C function and its driver, both together giving a ready to use and automatically generated CLI app.
Then, I'd like to share below those automatically generated web apps for a bunch of datasets. First, to illustrate the result; second, with the hope that at least some of them will be useful to someone. I also plan to update this article with new datasets little by little, and would be very happy to receive requests for datasets to process. Don't hesitate to contact me! Datasets below are in alphabetical order, with a link to the source. The percentage is the expected accuracy estimated from 10-fold cross validation.
|Predictor||Source||Expected accuracy||Web app|
|Heart disease detection||OpenML||82.96%||link|
|Iris type from petal and sepal dimensions||OpenML||97.33%||link|
|Leptograpsus crabs species from body dimensions||OpenML||99.50%||link|
|North-american mushroom edibility||OpenML||100.0%||link|
|Phoneme identification from harmonics||OpenML||85.80%||link|