I don't understand why it is allowed to use the information of soil peak humidity in the future but not the irrigation field. I mean one can approximate that value given the different peak in soil humidity.
Where was it specified that you can't use the irrigation field?
These there the exceptions below.
Having a solution that depends on other fields, or future weather patterns will not be useful for a farmer with one field or for someone who does not have reliable weather forecasting applications.
" We have also included an “irrigation” variables associated with each of the four fields. The irrigation variable is set to 1 when the irrigation is turned on and the soil moisture is rising and set to 0 when the irrigation is turned off." This tells us that irrigation is a context variable that has been added post-fact, so best practice would suggest that it not be used as it will not necessarily be available when productionalised. Also, the use case for the model is to help farmers plan their irrigation schedule, so having a model dependent on an existing future irrigation schedule is not useful?
A single model can be built leveraging the past irrigation pattern for the 4 different farm fields. Peak periods and irrigation patterns were provided to aid modelling. However, future data that wont be useful include things like weather considering the likely constraints stated by Zindi in the data section.
well they say this : " However, where the soil humidity peaks due to irrigation within the testset, you are provided with the peak soil humidity." and they said that you can't use data in the future and irregation information for the dates in the submission file is in fact in the future. @zindi please can you clarify here
You may not use any information from the future.
even the peak soil humidity ? I don't understand why these informatins are given in the test set and can't be used? why not just remove them ?
You may still use the peak soil humidity.
One thing that bother me with this is how the farmer will know this information ahead? and also why can we use this and not irrigatin field since if a farmer can give the peak information, it's way easier to give the planned irrigation for a given period. Am i getting something wrong
Thanks for this discussion. You may use peak humidity and irrigation patterns. See our final decision on this matter posted here: http://zindi.africa/competitions/wazihub-soil-moisture-prediction-challenge/discussions/496
thanks for the clarification