Bollinger Bands and their use in Stock Market Analysis (using Quandl & tidyverse in R)

1. Both the Stocks BOB and Canara shows similar price movement through out the year. both of them declined from oct 2016 to mid of Jan 2017 and then shown improvement till April. But After April both the stocks were on opposite directions. Canara went northwards, where as BOB went southward.
2. BOB shows W Bottoms in Jan 2017 and April followed with strong upward movement.

Prediction of Stock Prices

In this section, we will predict the prices for two selected bank PNB and Axis Bank. In stock market, generally the prices are dynamic and depends on various factors like news, weather, public policy, interest rate. It is difficult to predict the stock price behavior as it depends on lots of factor. In order to get more accuracy in prediction, we’ve used two different approach to come to prediction.

• Regression Approach
• ARIMA (Auto Regressive Intensive Moving Average)

In the last post, we have seen that the stock prices is also dependent on the traded quantity, but direction can be either ways. In our analysis, we will take consideration of these movements. We will also analyze the random part of the stock price movement, so called white noise and will include in our prediction model.

There is also available study on white noise on Analytics Vidhya by Tavish Srivastava.

The following points are the steps to arrive at Predictions

• Download the dataset in mothly format from Quandl
• Build the regresssion Model “m1” & “m3”
• Build the ARIMA model “p2.df” & “p4.df”
• Combine the regression and Arima model together “p1.df” & “p3.df”
• Prediction dataset has parameter like Close Price, Confidence intervals (“lwr(lower)” & “upr(Upper)”, Key (Prediciton/Actual)
• Add Date variable in the predcition dataset for three months “October”, “November” & “December”.
• Reorder the columns of Prediction Datasets.
• Modification of the Actual Datasets for Axis and PNB; keep only Date and Close price by using dplyr package.
• Add the variables like Confidence interval (“lwr” & “upr”), “Key”
• Combine Both the dataset Prediction and Actual into new dataframe like “Axis_com”
• Apply ggplot for the both the dataset
• use gridExtra package to arrange the plots.

Note: ggplot shows the prediction and actual prices. Predcition prices has the band for lower and upper limit.

    ## Download the data Set
PNB = Quandl("NSE/ICICIBANK",collapse="monthly",start_date="2016-09-01",type="raw")
Axis=Quandl("NSE/AXISBANK",collapse="monthly",start_date="2016-09-01",type="raw")

## Convert the PNB & Axis Data Set into df for regression model

PNB_df=PNB
Axis_df=Axis
colnames(PNB_df)<-c("Date","Open","High","Low","Last","Close","TTQ","Turnover")
colnames(Axis_df)<-c("Date","Open","High","Low","Last","Close","TTQ","Turnover")

## Change the scale of Trade quantity

PNB_df$TTQ<-PNB_df$TTQ/100000
Axis_df$TTQ<-Axis_df$TTQ/100000

## Regression models

m1=lm(PNB_df$Close~PNB_df$High+PNB_df$Low+PNB_df$TTQ)

p1.df=as.data.frame(predict(m1,interval="predict"))

m3=lm(Axis_df$Close~Axis_df$High+Axis_df$Low+Axis_df$TTQ)

p3.df=as.data.frame(predict(m3, interval="predict"))

## Forecast using ARIMA to take out the seasonality and cyclic part of the stock

m2=arima(diff(PNB_df$Close),order=c(1,0,0))
m4=arima(diff(Axis_df$Close),order=c(1,0,0))
p2.df=as.data.frame(predict(m2,n.ahead=3))
p4.df=as.data.frame(predict(m4,n.ahead=3))

## Combining the Random and Stock  together

p1.df=p1.df[1:3,]
p1.df$fit=p1.df$fit+p2.df$pred
p3.df=p3.df[1:3,]
p3.df$fit=p3.df$fit+p4.df$pred

## Create the date df for next three months

date<-as.data.frame(as.Date(c("2017-10-31","2017-11-30","2017-12-31")))
colnames(date)=c("date")

## Modify the predict dataset and add "key" variable for PNB

p1.df<-cbind(p1.df,date)
p1.df["Key"]<-"Predicted"
p1.df<-p1.df[,c("date","fit","lwr","upr","Key")]

## Modify the predict dataset for Axis and add variable "Key"

p3.df<-cbind(p3.df,date)
p3.df["Key"]<-"Predicted"
p3.df<-p3.df[,c("date","fit","lwr","upr","Key")]


## Rename the columns
colnames(p1.df)<-c("Date","Close","lwr","upr","Key")
colnames(p3.df)<-c("Date","Close","lwr","upr","Key")

## Modify the PNB_df dataset

PNB_df<-PNB%>%select("Date","Close")
Axis_df<-Axis%>%select("Date","Close")

## Add two variable for confidence interval "lwr" and "upr"
var<-c("lwr","upr")

PNB_df[var]<-NA
Axis_df[var]<-NA

## Add the Key variable for Actual data

PNB_df["Key"]<-"Actual"
Axis_df["Key"]<-"Actual"

## Rbind the predicted and actual value for both of the Stocks

PNB_com=rbind(PNB_df,p1.df)
PNB_com$Date<-as.Date(PNB_com$Date)

Axis_com=rbind(Axis_df,p3.df)
Axis_com$Date<-as.Date(Axis_com$Date)

## Visualisation

PNB_Plot<-ggplot(data=PNB_com,aes(x= Date, y = Close,color=Key,label=Close)) +
  # Prediction intervals
  geom_ribbon(aes(ymin = lwr, ymax = upr, fill = Key), 
              fill = "khaki2", size = 0)+
  geom_line(size = 1.7) + 
  geom_point(size = 2)+
  labs(title = "Actual and Predicted Price, PNB", x = "Date",y="Price") +
  theme(text = element_text(family = 'Gill Sans', color = "#444444",hjust=0.5)
        ,panel.background = element_rect(fill = "honeydew")
        ,panel.grid.minor = element_blank()
        ,panel.grid.major = element_blank()
        ,plot.title = element_text(size = 20,hjust=0.5,colour="orangered4")
        ,axis.title = element_text(size = 18, color = '#555555')
        ,axis.title.y = element_text(hjust=0.5,size=15)
        ,axis.title.x = element_text(hjust = 0.5,size=15))

Axis_Plot<- ggplot(data=Axis_com,aes(x= Date, y = Close,color=Key,label=Close)) +
  # Prediction intervals
  geom_ribbon(aes(ymin = lwr, ymax = upr, fill = Key), 
              fill = "khaki2", size = 0)+
  geom_line(size = 1.7) + 
  geom_point(size = 2)+
  labs(title = "Actual and Predicted Price, Axis Bank", x = "Date",y="Price") +
  theme(text = element_text(family = 'Gill Sans', color = "#444444",hjust=0.5)
        ,panel.background = element_rect(fill = "honeydew")
        ,panel.grid.minor = element_blank()
        ,panel.grid.major = element_blank()
        ,plot.title = element_text(size = 20,hjust=0.5,colour="orangered4")
        ,axis.title = element_text(size = 18, color = '#555555')
        ,axis.title.y = element_text(hjust=0.5,size=15)
        ,axis.title.x = element_text(hjust = 0.5,size=15))

## Plots

grid.arrange(PNB_Plot,Axis_Plot,ncol = 1, nrow = 2)

In this article, I have focused on Predictive Analysis of bank stocks. I have summarized a bit on Bollinger Bands, which probably is the most important topic in stock analysis. I have also walked you through the volatility of bank stocks and ways to see through this volatility.

This ends our journey of comparative analysis of stock market data. I hope it will help you to make your mark in the world of stocks. Good Luck!

Aritra Chatterjee is a professional in the field of Data Science and Operation Management having experience of more than 5 years. He aspires to develop skill in the field of Automation, Data Science and Machine Learning.

Learn, engage,compete, and get hired!