This project is to use mcrioblog information to predict people's personality. Here is an example for precdicting Agreeableness (A).
In this project, we use supervised deep learning technique to build the prediction model. As the input data is not image format, we use HDF5 to code the input data. And the hdf5FileGeneration4Caffe.m is used to generate the corresponding hdf5 files (this file is very useful, it can be used to other tasks). Also, the location texts shows the location of the data.

After training, the loss can be decreased to 0.03. And the model is psA_iter_10000.caffemodel.

The five mat files are features (128-dim) using the trained model for each personality trait.

If you are interested, please contact us.