AIFeynman: Attempt 1

#!/usr/bin/env python3

# -*- coding: utf-8 -*-

"""

Created on Mon May 24 16:53:40 2021

@author: rajeevgangal @VOIS, GBIS.

"""

from operator import index

from turtle import shape

from numpy import integer

import pandas as pd

from tables import Cols, Column

import aifeynman

import time

td = pd.read_csv("/home/rajeevgangal/myProjects/python/data/Telco-Customer-Churn.csv") # read the famous telecom churn dataset

tdpost= td[['Churn']] #separate response variable

td.infer_objects

td['TotalCharges']= pd.to_numeric(td['TotalCharges'], errors= 'coerce') # some fields are easily changeable to numeric

myobjlist= list(td.select_dtypes(include='object').columns) #list other columns that are objects/categorial

myobjlist.pop(0) #customerID not needed in object list for one-hot encoding

td= pd.get_dummies(td,columns=myobjlist,dummy_na=True,dtype=float) #its hot hot hot encoding or how to increase dimensionaly to feel curse of the same

#drop some columns with complementary redundant info

td.drop('Churn_No',inplace=True, axis=1)

td.drop('customerID',inplace=True, axis=1)

td.drop('Churn_nan',inplace=True, axis=1)

#interpolate some missing values

td= td.interpolate(method ='linear', limit_direction ='forward')

#aifeynman apparently needs a space separated file with no index and no headers, so lets write that

td.to_csv('/home/rajeevgangal/myProjects/python/example_data/tdclean.csv', sep=" ", header=False, index=False)

tdvars=td.columns.tolist

tic = time.perf_counter() #likely to be long running so time it.

print("Starting feynman search ", tic)

print(tdvars) #print which vars are being used, churn should be last col.

aifeynman.run_aifeynman("/home/rajeevgangal/myProjects/python/example_data/", "tdclean.csv", 60, "14ops.txt", polyfit_deg=3, NN_epochs=500, vars_name=tdvars, test_percentage=20)

toc = time.perf_counter()

timetaken= (toc-tic)/3600

print(" Ran feynman example in ", timetaken ," hrs")