This is a joke project. But in any joke there is truth. :) Goal of this project - check popular lottery 6 from 45. My hypothesis was that if this lottery is scam, then results of lottery will be predictable. I collected data from 3000 editions of the lottery, train neural network, and check it on next results of lottery. But in this experiment a don't found any correlations in results.
For training we need dataset with results of lottery in csv format. In other csv file fill last results of lottery. Important: fill blank fields with 0 (zeroes). Zeroes in fields needs because i did not invented how do this in lua. Openoffice does this task more fast then me. :)
Download dataset: https://www.dropbox.com/s/vhwc4eb87kdc236/6from45.csv?dl=0
Download example input data for test: https://www.dropbox.com/s/jb4af39aenxlnpw/6from45input.csv?dl=0
Put this csv files in same directory with lua files.
How to use:
-
Make own dataset or download from link above in csv format. Create dataset in t7 format with make-dataset.lua
-
Train network with train.lua . In my case network need 500000 train iteration. With cuda - ~1h on GTX 1070
-
Make input data in csv format. Put in csv file last results from lottery 6 from 45.
-
Do predict lottery result with predict.lua
-
PROFIT :)
NOTE: starting all scripts with -h key will display short help
Good luck.