Predicting Future Yield with unknown future X_train values

[mehathab96]

Hello,

My name is Mehathab, I am an aspiring Data Scientist,
I have gone through your paper and found it very interesting,
I have made a new model using LSTM, GRU and CNN for the dataset, but now cannot understand how do we predict the future yield without the future x_train values?

also what is the sequencing you are using for the dataset creating,
what i am doing is,
my model will consider the last 10 data points for predicting yield,

Code i am using is :

def split_sequences(sequences, n_steps_in, n_steps_out):
    X, y = list(), list()
    for i in range(len(sequences)):
        # find the end of this pattern
        end_ix = i + n_steps_in
        out_end_ix = end_ix + n_steps_out - 1
        # check if we are beyond the dataset
        if out_end_ix > len(sequences):
            break
        # gather input and output parts of the pattern
        seq_x, seq_y = sequences[i:end_ix, 1:], sequences[end_ix - 1:out_end_ix, 0]
        X.append(seq_x)
        y.append(seq_y)
    return np.array(X), np.array(y)

where n_steps_in = 10, n_step_out = 5

since i am a new to Forecasting problems,
can you help me with this ?
is this correct?

[rongtongxueya]

same question.Personally, I think it is necessary to predict the year with the labeled data, so as to judge the accuracy of the prediction? If you have an answer, please let me know

[mehathab96]

same question.Personally, I think it is necessary to predict the year with the labeled data, so as to judge the accuracy of the prediction? If you have an answer, please let me know

I think what could be done to is use the last 'n' data points to predict the future yield,
for example using data from 2014-2018 to predict the yield on 2019
this can be achieved by time-lagged shift sequencing
( the assumption is that the yield produced in 2019 is directly dependent upon the data from 2014-2018 including the yield)

but for model creation and validation we need labelled data itself, once the model is perfected in a way, we could use this to predict the future values without having the actual future values

if you find a better approach/solution please comment