0
$\begingroup$

I have dataset (20 samples) containing timeseries on temperature, humidity etc (a total of 6 variables). Each timeseries is 5 complete days, which is 24 * 5 = 120 values.

So dataset has hourly values of temperatures etc but when it comes to lets say yields parameters I have only one vector containing subjective crop quality (scale 1-10), vitamin A, C, E content, protein content and fat content measured after the end of the crop.

I am familiar with machine learning using scikit-learn or tensorflow. But I am not sure how to represent this problem. The tricky part here for me is that I have just six values of yield per one sample.

I have read some articles and I have applied RNN (like in those articles) but they usually had some 'yields parameters' measured during plant growing.

How do I approach this?

I have thought about 6 models of RNN each for all yield parameters or maybe one RNN where last Dense layer is 6.

So far I transformed the data and it has shape: (20, 6, 120) and the y shape: (20, 6) or (20,1) (depends on approach).

$\endgroup$
4
  • $\begingroup$ Try reforming your data so that you have a single feature vector per target variable value. Neural networks will probably perform poorly in your case, you have too little data. $\endgroup$ Aug 22, 2022 at 11:58
  • $\begingroup$ @JanŠimbera you mean one vector of temperature to predict (for example) Vitamin A? $\endgroup$
    – suziex
    Aug 22, 2022 at 19:58
  • $\begingroup$ I struggle to understand what you are trying to predict - but if you are trying to predict Vitamin A contents, you should have a single vector of values per each Vitamin A measurement. $\endgroup$ Aug 23, 2022 at 13:14
  • $\begingroup$ @JanŠimbera hello again, one sample of X variable is timeseries of: temperature, humidity, visible light intensity, sound intensity, water level and crop ventilation power (total of six timeseries) and for this one sample my Y(what I have to predict) are subjective cultivation quality, Vitamin A, Vitamin C, Vitamin E, Protein and Fat (those are not timeseries, just single values) $\endgroup$
    – suziex
    Aug 23, 2022 at 16:54

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.