STF_RNN.py

from utils import *
import numpy as np
from models import stf_rnn
from keras.utils.np_utils import to_categorical
from keras.callbacks import ModelCheckpoint, EarlyStopping
import argparse

if __name__ == '__main__':

    parser = argparse.ArgumentParser(description='STF-RNN: Space Time Features-based Recurrent Neural Network for Predicting People Next Location')
    parser.add_argument('-e1',  dest='emb_size1', type=int, default=100, help='The space embedding size.')
    parser.add_argument('-e2',  dest='emb_size2', type=int, default=6, help='The time embedding size.')
    parser.add_argument('-r',  dest='rnn_size', type=int, default=20, help='The recurrent hidden size.')
    parser.add_argument('-w',  dest='window_size', type=int, default=2, help='The window size.')
    parser.add_argument('-t',  dest='min_time', type=int, default=1800, help='The time threshold in seconds.')
    parser.add_argument('-d',  dest='min_dist', type=float, default=0.2, help='The distance threshold.')
    parser.add_argument('-e',  dest='epochs', type=int, default=100, help='The number of training epochs.')
    parser.add_argument('-b',  dest='batch_size', type=int, default=30, help='The mini batch size.')

    args = parser.parse_args()

    print(args)

    acc = []
    window_size = args.window_size
    k = 0
    for user_id in range(183):

        print("Processing user %d." % user_id)
        try:
            data_path = 'Data/%03d/Trajectory/*.plt' % user_id
            traj = read_data(data_path)

            s_inputs, t_inputs, outputs, N = prpare_data(traj,
                                                        window_size=window_size,
                                                        min_dist=args.min_dist,
                                                        min_time=args.min_time)

            if N == 0 or len(s_inputs) <= 8:
                continue
            outputs_ = to_categorical(outputs, N)
            s_inputs = np.array(s_inputs).astype(np.int64)
            t_inputs = np.array(t_inputs).astype(np.int64)
            model = stf_rnn(N, args.emb_size1, 24, args.emb_size2, window_size, args.rnn_size)
            #print(model.summary())
        
            model_path = 'stf_rnn_models/model_u%d.h5' % user_id
            call_backs = [ModelCheckpoint(model_path, monitor='val_acc', verbose=1, save_best_only=True), 
                    EarlyStopping(monitor='val_acc', patience=8, verbose=1, mode='auto')]

            hist = model.fit([s_inputs, t_inputs], outputs_,
                            validation_split=0.2,
                            epochs=args.epochs,
                            batch_size=args.batch_size,
                            verbose=1,
                           callbacks=call_backs)

            acc.append(max(hist.history['val_acc']))
            k += 1
        except :
            pass

    
    print(np.mean(acc))