|
| 1 | +#%% |
| 2 | +importwarnings |
| 3 | +warnings.filterwarnings('ignore') |
| 4 | + |
| 5 | +importos |
| 6 | +importnumpyasnp |
| 7 | +importpandasaspd |
| 8 | +importscipy.statsasst |
| 9 | +importstatsmodels.apiassm |
| 10 | +importmatplotlib.pyplotasplt |
| 11 | + |
| 12 | +fromsklearnimportlinear_model |
| 13 | +fromsklearn.metricsimportmean_squared_error,r2_score,confusion_matrix,accuracy_score |
| 14 | +fromsklearn.model_selectionimporttrain_test_split |
| 15 | +fromsklearn.metricsimportroc_curve |
| 16 | +fromsklearn.metricsimportroc_auc_score |
| 17 | +# %% |
| 18 | +os.chdir('D:/machine_learning/DATA_SET/credit_risk') |
| 19 | +df_a=pd.read_csv('credit_risk.csv') |
| 20 | +df_a.head() |
| 21 | +# %% |
| 22 | +df_a.isnull().sum() |
| 23 | +# %% |
| 24 | +df_a.dropna(axis=0,inplace=True) |
| 25 | +df_a.isnull().sum() |
| 26 | +# %% |
| 27 | +df_a.dtypes |
| 28 | +# %% |
| 29 | +df_a["CoapplicantIncome"]=df_a["CoapplicantIncome"].astype('float') |
| 30 | +df_a["ApplicantIncome"]=df_a["ApplicantIncome"].astype('float') |
| 31 | +print("AFTER CHANGING INT & OBJECT TO FLOAT") |
| 32 | +df_a.dtypes |
| 33 | +# %% |
| 34 | +df_a["Dependents"].value_counts() |
| 35 | +# %% |
| 36 | +cleanup_nums= { |
| 37 | +"Married": {"Yes":1,"No":0}, |
| 38 | +"Self_Employed": {"Yes":1,"No":0}, |
| 39 | +"Property_Area": {"Rural":0,"Urban":1,"Semiurban":2}, |
| 40 | +"Education": {"Graduate":1,"Not Graduate":0}, |
| 41 | +"Loan_Status": {"Y":1,"N":0}, |
| 42 | +"Gender": {"Male":1,"Female":0}, |
| 43 | +"Dependents": {"0":0,"1":1,"2":2,"3+":3}, |
| 44 | +} |
| 45 | + |
| 46 | +df_a.replace(cleanup_nums,inplace=True) |
| 47 | +df_a.head() |
| 48 | +# %% |
| 49 | +df_corr=df_a[['Gender','Married','Dependents','Education', |
| 50 | +'Self_Employed','ApplicantIncome','CoapplicantIncome','LoanAmount', |
| 51 | +'Loan_Amount_Term','Credit_History','Property_Area','Loan_Status']] |
| 52 | + |
| 53 | +df_corr.head() |
| 54 | +# %% |
| 55 | +plt.imshow(df_corr.corr(),cmap=plt.cm.Blues ,interpolation='nearest') |
| 56 | +cmap='coolwarm' |
| 57 | +plt.colorbar() |
| 58 | +tick_marks= [iforiinrange(len(df_corr.columns))] |
| 59 | +plt.xticks(tick_marks,df_corr.columns,rotation='vertical') |
| 60 | +plt.yticks(tick_marks,df_corr.columns) |
| 61 | +plt.show() |
| 62 | +#%% |
| 63 | +boolean_col='Loan_Status' |
| 64 | +cols= ['Gender','Married','Dependents','Education','Self_Employed', |
| 65 | +'ApplicantIncome','CoapplicantIncome','Credit_History' |
| 66 | + ] |
| 67 | + |
| 68 | +xTrain=df_a[cols].values |
| 69 | +yTrain=df_a[boolean_col].values |
| 70 | +# %% |
| 71 | +st.chisqprob=lambdachisq,df:st.chi2.sf(chisq,df) |
| 72 | + |
| 73 | +model=sm.Logit(yTrain,xTrain ) |
| 74 | +result=model.fit() |
| 75 | +result.summary(xname=cols,yname=boolean_col,title='Logit Model',alpha=1) |
| 76 | +# %% |
| 77 | +defcalculate_accuracy(predictions,real): |
| 78 | +correct=0 |
| 79 | +foriinrange(len(predictions)): |
| 80 | +ifround(predictions[i])==round(real[i]): |
| 81 | +correct+=1 |
| 82 | +returncorrect*1.0/len(predictions) |
| 83 | +# %% |
| 84 | +train_predictions=result.predict(xTrain) |
| 85 | + |
| 86 | +train_accuracy=calculate_accuracy(train_predictions,yTrain ) |
| 87 | +print("Train Accuracy: ",train_accuracy*100 ) |
| 88 | +# %% |
| 89 | +train_predictions= (train_predictions>0.5) |
| 90 | + |
| 91 | +train_cm=confusion_matrix(yTrain,train_predictions,labels= [1.0,0.0]) |
| 92 | +print(train_cm ) |
| 93 | +# %% |
| 94 | +labels= ['0','1'] |
| 95 | +cm=train_cm |
| 96 | +fig=plt.figure() |
| 97 | +ax=fig.add_subplot(111) |
| 98 | +cax=ax.matshow(cm,cmap='viridis') |
| 99 | +plt.title('Confusion matrix') |
| 100 | +fig.colorbar(cax) |
| 101 | +ax.set_xticklabels(['']+labels) |
| 102 | +ax.set_yticklabels(['']+labels) |
| 103 | + |
| 104 | +r=0 |
| 105 | +c=0 |
| 106 | +forlistItemincm: |
| 107 | +forcellIteminlistItem: |
| 108 | +ax.text(c,r,cellItem,va='center',ha='center',color='r') |
| 109 | +c+=1 |
| 110 | +c=0 |
| 111 | +r+=1 |
| 112 | + |
| 113 | +plt.xlabel('Actual label') |
| 114 | +plt.ylabel('Predicted label') |
| 115 | +plt.show() |
| 116 | +# %% |