Using RapidMiner Neural Net Model with JavaScript in a smartwatch

klemens1959

I have developed a RM Neural net (NN) model which predicts ADLs (activities of daily living) based on smartwatch sensor data. The model works very well and makes good predictions. I want now to bring back the NN model to my smartwatch for real time predictions. As the model is contained/described in an xml file my idea is to extract the necessary parameters from the file and use those parameters in a JavaScript function to compute the class associated with the real time data (Samsung Gear uses Tizen/HTML5 and JavaScript internally for apps). Is there a description around which can tell me how I can use the model data for that, esp. which math operations are used to compute the prediction class. I think this cannot be too complicated as NN predictions basically is adding and weighting.

Thanks for ideas!
Klemens

JEdward

I had a look to see if PMML can be run on Javascript and stumbled onto Encog. 

I'm not sure how this would perform, but someone has coded Neural Networks into Javascript. 
http://www.codeproject.com/Articles/477689/JavaScript-Machine-Learning-and-Neural-Networks-wi

http://www.heatonresearch.com/encog/

Have a look and see if it suits.