# unit 2.4 logistic regression

import numpy as np

import numpy.random as rnd

import pandas as pd

from statsmodels.api import Logit as smLogit

from sklearn.cross_validation import KFold

#from sklearn.cross_validation import train_test_split as ttsplit

import matplotlib.pyplot as plt

import re

import os

from svm_predict import plot_predict, do_boxplot

def get_app_title():

"get app title"

return 'Logit Regression'

def get_app_file():

"get app file prefix"

return 'lr_sm_'

def make_plotdir():

return plotdir

def read_data():

return loansData

def load_data():

return loansData, testData

# don't need to split dep, indep vars here

# return loans_df, loans_y, test_df, test_y, numeric_vars

# logistic regression example

# Q: What is the probability of getting a loan from the Lending Club

# for $10,000 at an interest rate <= 12% with a FICO score of 750?

# Model: log(p/1-p) = mx + b "logit function"

# invert to find p(x) = 1 / (1 + exp(mx + b)) "logistic function"

# Wiki: The logit function is the quantile function of the

# logistic distribution, while the probit is the quantile function

# of the normal distribution. See https://en.wikipedia.org/wiki/Logit

def do_logit(loansData, indep_variables, print_out=False):

return result

def fit_score_logit(loansData, numeric_vars, cutoff=0.5):

return score, result, loansData

def logistic_fn_orig(loanAmount, fico, params):

return p

# Why do coefficients have opposite sign? smLogit or IR_TF lambda is backwards?

def logistic_fn(loanAmount, fico, params):

return p

# Probability isn't binary, try assumption: p<70% won't get the loan.

# if p >= 0.70 then 1, else 0

def pred_orig(loanAmount, fico, params):

return msg

def pred(loanAmount, fico, params):

return msg

def test_results(result):

print('\nThe probability that we can obtain a loan at less than 12 percent interest for $10000 USD with a FICO score of 720 is: %.1f percent. It is more likely than not we will get the loan for under 12 percent.' % ( 100 - 100 * logistic_fn(10000, 720, result.params) ))

def plot_fico_logit(result, plotdir):

plt.savefig(plotdir+'fico_logistic.png')

def plot_loan_logit(result, plotdir):

plt.savefig(plotdir+'loan_logistic.png')

def plot_loan_fico(loansData, result, plotdir):

plt.savefig(plotdir+'loan_v_fico.png')

def set_plot_predict(plotdir, app, appf, label, indep_vars, full_df):

plot_predict(plotdir, app, appf, label, indep_vars, test_df, test_y, pred_y)

def check_cutoff(dframe, vlist):

print(">>> opt cutoff %.2f, score %.5f, vars=%s" % (opt_cutoff, opt_score, vlist))

def calc_score(dframe, params, cutoff=0.5):

return score, dframe

def fit_train_score(vlist, dframe):

"fit training data, get train score"

tr_score, result, dframe = fit_score_logit(dframe, vlist, cutoff=0.5)

return tr_score, result

def random_opt(varlist, init_list, dframe, print_out=False):

return score, vlist, result, scores

def run_opt(numeric_vars, dframe, app, appf, plotdir):

return opt_score, opt_list, opt_result, opt_scores

def fit_predict(loansData, validData, indep_vars):

return score, result, validData

def do_kfold_cv(loansData, indep_vars, n_folds=5, print_out=False):

return newpar, scores

# main program

def main():

2 * np.std(scores), score))

if __name__ == '__main__':

main()