Use cell arrays with neural networks in a regression problem.

I want to use Neural Networks to predict a numeric value using text and numerical data. Here is an example:

It's a regression problem. If I try to use the Neural fitting (nftool) app, only allows use numerical data.

A solution that I can think of is to assign numerical values to each class in the text data values , but I think it is too computationally expensive , as my dataset has a size 17000x40.

Is there any way to use neural networks in a regression problem using cell data type ?

Answer 1

[Sean de Wolski]

First, convert your cell array to a table (cell2table). Tables make working with this type of data much easier.

Then, convert the cell strings to categoricals and then dummy code them using dummyvar or my dummytable, attached.

x = table(rand(3,1),{'A';'B';'C'});
x.Var2 = categorical(x.Var2);
xdummy = dummytable(x);
Xnumeric = table2array(xdummy)

[Manuel Valenciano]

I have a problem with the categorical function. Some variables of my problem are of the following type:

I want to convert it to a categorical array but the categorical function gives me the following error:

Error using categorical (line 321)
Could not find unique values in DATA using the UNIQUE function.

Caused by:
    Error using cell/unique (line 85)
    Input A must be a cell array of strings.

[Sean de Wolski]

You'll need to convert the numeric values to strings first.