I have a dataset which is of the dimensions nxm. Columns 1-m are the features which I would like to use, and the last column i.e. column m consists of labels/targets (k different classes, string type). The first column is of the DateTime datatype format, so I am not sure if I need to convert it/typecast it to a float or some other datatype. I would like to use Recurrent Neural Networks (RNN) for multi-class classification, but the thing is that RNNs take sequential data in general.
I am confused as to the particular shape or format in which I need to provide the data to the RNN for multi-class classification. To be precise, I would like to know that is there any specific way which would be ideal, such as creating a list of lists with each list within containing the features or some different method etc. Any help in this regard is highly appreciated. Thanks!
Dear Chatterjee,
I don't understand why you insist on using Recurrent Neural Network for tabular data. Your problem is a classical classification problem. All you have to do is convert your (non-numeric) data to numeric data. For example, in the case date time you can create more features from it ( number of second, day, Week of month, month of year ......) . Once you cleaned your table, you could choose and test a model from the plethora of models and choose the best one that fit your performance metric.
Hi Joyjit,
Following tutorials can be helpful:
https://machinelearningmastery.com/multi-class-classification-tutorial-keras-deep-learning-library/
https://machinelearningmastery.com/sequence-classification-lstm-recurrent-neural-networks-python-keras/
Given that, you have time based data, you might find some regression based problems in following index, which can be improvised to adapt for classification:
https://machinelearningmastery.com/category/lstm/
- Badges
- Science topic