Machine Learning - Logistic Regression Model(Imarticus Case Study)

library(readxl)
attrition <- read_excel("C:/Users/Hxtreme/Desktop/attrition.xlsx")

my_attrition=attrition

The main data set is been swaped to a Sub data set.

View(my_attrition)

VIEW: To display the entire Data Set.

str(my_attrition)

Classes ‘tbl_df’, ‘tbl’ and 'data.frame':   160 obs. of  9 variables:
 $ CustID    : num  101901 102056 102522 103149 103866 ...
 $ Gender    : num  1 0 1 1 0 0 1 1 0 1 ...
 $ Age       : num  30 17 54 42 30 23 28 19 48 40 ...
 $ Income    : num  0 1 1 1 1 0 1 1 0 1 ...
 $ FamilySize: num  5 1 4 2 2 4 2 5 4 5 ...
 $ Education : num  20 12 18 17 12 16 18 16 15 16 ...
 $ Calls     : num  37 25 48 51 26 18 29 28 16 31 ...
 $ Visits    : num  3 1 3 2 1 0 2 1 3 3 ...
 $ Churn     : num  1 0 1 1 0 0 1 1 1 1 ...

STR(Structure):It is a compact way to display the structure of an R object.

summary(my_attrition)

     CustID           Gender            Age            Income         FamilySize   
 Min.   :101901   Min.   :0.0000   Min.   :17.00   Min.   :0.0000   Min.   :1.000  
 1st Qu.:126719   1st Qu.:0.0000   1st Qu.:22.00   1st Qu.:0.0000   1st Qu.:2.000  
 Median :151060   Median :1.0000   Median :31.00   Median :1.0000   Median :3.000  
 Mean   :151144   Mean   :0.5813   Mean   :35.67   Mean   :0.5062   Mean   :3.131  
 3rd Qu.:176099   3rd Qu.:1.0000   3rd Qu.:46.00   3rd Qu.:1.0000   3rd Qu.:4.000  
 Max.   :199131   Max.   :1.0000   Max.   :82.00   Max.   :1.0000   Max.   :5.000  
   Education         Calls           Visits          Churn       
 Min.   :12.00   Min.   : 3.00   Min.   :0.000   Min.   :0.0000  
 1st Qu.:12.00   1st Qu.:15.75   1st Qu.:1.000   1st Qu.:0.0000  
 Median :14.00   Median :22.00   Median :2.000   Median :1.0000  
 Mean   :14.96   Mean   :25.22   Mean   :1.906   Mean   :0.5312  
 3rd Qu.:17.00   3rd Qu.:32.00   3rd Qu.:3.000   3rd Qu.:1.0000  
 Max.   :20.00   Max.   :65.00   Max.   :5.000   Max.   :1.0000  

SUMMARY : It is a generic function used to produce result summaries of the results of various model fitting functions.

my_attrition=my_attrition[,-1]

As no need of cUSTID, I have removed it.

my_attrition$Churn=as.factor(my_attrition$Churn)

my_attrition$Gender=as.factor(my_attrition$Gender)

my_attrition$Income=as.factor(my_attrition$Income)

FACTOR : are the data objects which are used to categorize the data and store it as levels. They can store both strings and integers. They are useful in the columns which have a limited number of unique values.

library(ggplot2)

GGPLOT2 : The ggplot2 package offers a powerful graphics language for creating elegant and complex plots.

ggplot(my_attrition,aes(x=Income,y=Churn,col=Churn))+geom_jitter()+labs(title = "INCOME OF CHURN")

The above plot deals with the income of the churn. GEOM_JITTER : It adds a small amount of random variation to the location of each point, and is a useful way of handling overplotting caused by discreteness in smaller datasets.

ggplot(my_attrition,aes(Age,Calls))+stat_density2d(geom = "tile",aes(fill=..density..),contour = FALSE)+geom_point(colour="white")+labs(title = " 2D Plot ")

The above dispaly a Good deference and elaborate the variables in 2D plot.

COUNTOUR : is a generic function with only a default method in base R. … For more control, add contours to a plot, or add axes and frame to a contour plot.

library("scatterplot3d")

SCATTERPLOT : a graph in which the values of two variables are plotted along two axes, the pattern of the resulting points revealing any correlation present.

scatterplot3d(my_attrition[,1:3],
              main="3D Scatter Plot",
              xlab = "Education (cm)",
              ylab = "Income (cm)",
              zlab = "Age (cm)")

Unknown or uninitialised column: 'color'.

SCATTERPLOT3D : for the supplementary points or regression planes into an already generated graphic.

library(caTools)

CATOOLS : Contains several basic utility functions including: moving (rolling, running) window statistic functions, read/write for GIF and ENVI binary files, fast calculation of AUC, LogitBoost classifier, base64 encoder/decoder, round-off-error-free sum and cumsum, etc.

SAmple=sample.split(my_attrition,SplitRatio = 0.8)

SAmple

[1]  TRUE  TRUE  TRUE  TRUE  TRUE FALSE FALSE  TRUE

SAMPLE.SPLIT : is used to divide the data into two sets….. train set and test set. Below piece of code is used to divide the data into train and test set.

Train=subset(my_attrition,SAmple=="TRUE")

Length of logical index must be 1 or 160, not 8

Train

Train : We can divide data into a particular ratio here it is 80% train and we train the machine.

Test=subset(my_attrition,SAmple=="FALSE")

Length of logical index must be 1 or 160, not 8

Test

We can divide data into a particular ratio here it is 20% in a test dataset.

my_attrition_eq=glm(Churn~.,Train,family="binomial")

my_attrition_eq

Call:  glm(formula = Churn ~ ., family = "binomial", data = Train)

Coefficients:
(Intercept)      Gender1          Age      Income1   FamilySize    Education        Calls  
   -6.20407      1.16110     -0.02459      0.85539      0.58304      0.20786      0.02800  
     Visits  
    0.35617  

Degrees of Freedom: 119 Total (i.e. Null);  112 Residual
Null Deviance:      164.2 
Residual Deviance: 126.2    AIC: 142.2

We obatain model to find accuray,here we also check AIC & Residual deviance.

AIC : The Akaike Information Criterion (AIC) provides a method for assessing the quality of your model through comparison of related models. It’s based on the Deviance, but penalizes you for making the model more complicated.

Residual deviance : The Residual Deviance has reduced by 22.46 with a loss of two degrees of freedom.

Test_pred=predict(my_attrition_eq,Test,type="response")

Test_pred

        1         2         3         4         5         6         7         8         9 
0.3525772 0.8257432 0.6993325 0.4568739 0.8009582 0.7789753 0.5361129 0.9024260 0.8756390 
       10        11        12        13        14        15        16        17        18 
0.8638592 0.3061531 0.5834685 0.6953019 0.7384999 0.7844095 0.1266510 0.5755652 0.4039113 
       19        20        21        22        23        24        25        26        27 
0.3725526 0.4299794 0.6883279 0.1810669 0.0806694 0.7729659 0.7611589 0.6509631 0.2653015 
       28        29        30        31        32        33        34        35        36 
0.6293396 0.5150554 0.3710720 0.8393425 0.1324726 0.2440157 0.1660396 0.9607420 0.4591172 
       37        38        39        40 
0.6796177 0.1493302 0.4410668 0.2303010 

Predict function is used to prediction . Type = RESPONSE :The type=“response” option tells R to output probabilities of the form P(Y = 1|X) , as opposed to other information such as the logit . If no data set is supplied to the predict() function, then the probabilities are computed for the training data that was used to fit the logistic regression model.

library(ROCR)

ROCR (with obvious pronounciation) is for evaluating and visualizing classifier performance. It is… …easy to use: adds only three new commands to R. …flexible: integrates tightly with R’s built-in graphics facilities.

ROCRPred=prediction(Test_pred,Test$Churn)

ROCR_Perfor=performance(ROCRPred,"acc")

plot(ROCR_Perfor)

Before the prediction, ROCR shows the high streak were as that is the accuracy.

T=table(actualvalue=Test$Churn,predictedvalue=Test_pred>0.5)

T

           predictedvalue
actualvalue FALSE TRUE
          0    17    6
          1     1   16

More generally in binary classification, a false positive is an error in data reporting in which a test result improperly indicates presence of a condition, such as a disease (the result is positive), when in reality it is not present, while a false negative is an error in which a test result. TO cAL ACC = The false positive rate is calculated as b ÷ (b + d), or 1 – the specificity; the true positive rate (sensitivity) as a ÷ (a + c); the true negative rate (specificity) as d ÷ (b + d), or 1 – the false positive rate; and the positive predictive value as a ÷ (a + b).

sum(diag(T))/sum(T)

[1] 0.825

As per the Accuracy i prove that my analysics is 83% Accurate.