I have started doing a project named Talking Data User Demographics. In this project i need to find the gender and age group of a person whose phone data is provided to me. In the datset given to me i could find only phone model and phone brand for almost 70% of the data and remaining 30% has events,event labels, location,many other features includeing model and brand. So, i thought of building a model seperately for devices with events and devices without evnts. As a part of finding correct model i have used various algorithms. When i tried neural nets i have got less log loss for devices with out events so, i thought of trying the same for devices with events also but it is giving the error "sparse matrices are not supported." But if it is so it is runnning fine when i have the model for devices with out events. Below i have attached a snippet as a proof of what i have told, please help me in understanding this any suggestion would be appreciated.
The file attached contain a code which runs the code perfectly even when sparse matrices are used in model.fit().[imagecan be seen here] (https://i.stack.imgur.com/0I99t.png).
Now in the below attached file it is giving the error when sparse matrix are passed to model.fit() [image can be seen here] (https://i.stack.imgur.com/E1b9f.png)
please gothrough the entire code for best reference.[code for reference] (https://drive.google.com/file/d/1hR5yLt8nLyde8CVbKBLBy35ghLJ_7Lxt/view)