Reshaping TrainX and trainY in LSTM RNN Project

Question

I am working on multivariate time based LSTM RNN Project. Because it is a multivariate, my goal is to train the model with multiple parameters and as an output I want to get predictions for every parameter in the dataset. I will add all the code and information below.

Here is the data set

Time(sec)   Lat(deg)    Lon(deg)    Alt(m)  Yaw(deg)    Pitch(deg)  Roll(deg)   Vx(m / sec) Vy(m / sec) Vz(m / sec)
0   1.00    42.461525   33.000000   2802.544922 0.000000    0.000000    0.000000    150.000000  0.000000    -0.000000
1   1.02    42.461525   33.000000   2802.544922 0.000000    0.000000    0.000000    150.000000  0.000000    -0.000000
2   1.04    42.461579   33.000000   2802.544922 0.000000    0.000000    0.000000    150.008987  0.000000    -0.000000
3   1.06    42.461579   33.000000   2802.544922 0.000000    0.000000    0.000000    150.008987  0.000000    -0.000000
4   1.08    42.461633   33.000000   2802.544922 0.000000    0.000000    0.000000    150.035843  0.000000    -0.000000

and here is the reshaping code for that data set:

trainX = [] trainY = []
    
n_future = 1   
n_past = 100  
    

 for i in range(n_past, len(df_missile_training_scaled) - n_future +1):
    trainX.append(df_missile_training_scaled[i - n_past:i, 0:df_missile_training_scaled.shape[1]])
    trainY.append(df_missile_training_scaled[i + n_future - 1:i + n_future, 0])

with this code i am reshaping the input as I wanted but it only outputs first column "Lat(deg)" because of the shape of the trainY.

trainX, trainY = np.array(trainX), np.array(trainY)
print('trainX shape == {}.'.format(trainX.shape))
print('trainY shape == {}.'.format(trainY.shape))

Outputs;

trainX shape == (171900, 100, 9).
trainY shape == (171900, 1).

My goal is to change it to a shape that after the training, model can predict multiple parameters (columns) on one training.

I tried this code but some how it gives me;

trainY shape == (171900, 1).

Here is the code I tried:

trainX = []
trainY = []

n_future = 1
n_past = 100

    for i in range(n_past, len(df_missile_training_scaled) - n_future +1):
        trainX.append(df_missile_training_scaled[i - n_past:i, 0:df_missile_training_scaled.shape[1]])
        trainY.append(df_missile_training_scaled[i + n_future:df_missile_training_scaled.shape[1]])

My question is how can I do this reshaping properly. And what I'm aiming for can be done for the neural network model?

I hope I was able to explain my question. If there is something I missed or couldn't explain, I can add it. Thanks in advance for reading

Reshaping TrainX and trainY in LSTM RNN Project

0 Answers0