I have eight columns in my data with 3900 rows, column 1 is time (with 1-day
interval) and the seven columns are the parameters related to car price and the ninth column represents sales of car and I am trying to forecast the number
of sales for future time steps. Help me to set the number of samples, time steps and
features in this data for an LSTM?
Based on your provided information:
The samples are simply equal to the number of data. In your case, there are 3900.
And features are constructed based on your columns as well.
Once you organize your training samples, it is needed to convert them into the flattened array. (e.g. an image with 28*28 pixels will be converted into a vector with 784-dimensions). Ultimately, according to the batch size, its time to feed them into the neural network.
Let me just add that, I reckon due to the low training sample, the network might not lead to the great prediction.
Perhaps this will help: https://www.datacamp.com/community/tutorials/lstm-python-stock-market
- Badges
- Science topic