#Abstract
##The model is based on Embedding+MLP, in MLP the model combines 3 parallel sub-model(DNN, CIN, AutoInt)
##The 3 parallel parts share the same embedded input, and use a combination layer to get the CTR.(pretty much like the xDeepFM model, this model is actually a fork of it)

#Code
##Code of this model is written and run in jupyter notebook, build the model requries only tensorflow2.1.0 and numpy(yeah, nothing else!!!! But the whole project need some other lib such as pandas and matplotlib)
##And to run the run and simply analysis the results requires some commonly seen lib, they are all listed in the head of the code file.

#Dataset
##Dataset of this model comes from here: https://tianchi.aliyun.com/dataset/dataDetail?dataId=56
##After you donwload the dataset from the url, you need some basic data dig(I used pandas)
##Short story, you need your dataset feeding to model looks like this:(here are some .csv examples)

label,cms_segid,cms_group_id,final_gender_code,age_level,pvalue_level,shopping_level,occupation,new_user_class_level,cate_id,campaign_id,customer,brand,price
0,0,10,1,4,0.0,3,0,1.0,6281,19451,96837,259125.0,99.0
0,7,2,2,2,1.0,3,0,2.0,4301,248040,202257,339499.0,129.0
0,0,2,2,2,0.0,3,0,0.0,1540,408389,63260,393737.0,38.0
0,0,3,2,3,2.0,3,0,0.0,5993,126962,156852,257537.0,15.0
0,0,4,2,4,0.0,3,0,3.0,24,338688,18656,0.0,5.3
......

##there is a dataProcess.py file, it's what I used to dig the data, only as a reference, cannot be directly use!!

###the code is not using any GPU computation(don't ask, I'm too poor and using AMD and haven't friguring out how to run tf on AMD gpus....