Predictive Analytic Dashboard.Rmd

---
title: "ANZ"
output: 
  flexdashboard::flex_dashboard:
    orientation: columns
    vertical_layout: fill
---

```{r setup, include=FALSE}
library(flexdashboard)
library(ggplot2)
library(plotly)
library(dplyr)
library(readxl)
library(fpp2)
```

```{r}
data<-read_excel("ANZ_synthesised_transaction_dataset.xlsx")
```

Predictive Analytics
==================================================================

Column {data-width=550}
-----------------------------------------------------------------------

### Histogram of customers' salary: Most of the customers paid by RM1000-1200

```{r}
data1<- data%>%filter(txn_description=="PAY/SALARY")
plot_ly(data1, x=~amount, type="histogram")
```


### Relationship between salary and age: Salary decreased by age

```{r}
d1<-ggplot(data1, aes(x=age,y=amount))+geom_point()+geom_smooth(method = "lm",se=FALSE)
ggplotly(d1)
```

Column {data-width=450}
-----------------------------------------------------------------------

### Relationship between salary and gender : Male get paid more than female

```{r}
d2<-ggplot(data1, aes(x=gender,y=amount))+geom_boxplot()
ggplotly(d2)
```

### Relationship between salary and day : Most of the customers get paid on Monday
>1:Monday, 2:Tuesday, 3:Wednesday, 4:Thursday, 5:Friday, 6:Saturday, 7:Sunday

```{r}
d3<-ggplot(data1, aes(Weekday))+geom_bar()
ggplotly(d3)
```

Multiple Linear Regression Model
=====================================================================

Column {data-width=450}
-----------------------------------------------------------------------

### Model
>Since the p-value for gender and age are small, so the test is significant that gender and age are significantly effect the amount of salary.

```{r}
d4<-lm(amount~gender+age+Weekday,data1)
summary(d4)
```

Column {data-width=550}
-----------------------------------------------------------------------


### Residuals diagnostic(Autocorrelation test)

```{r}
res<-residuals(d4)
d5<-ggAcf(res)
ggplotly(d5)
```

### Residuals diagnostic(Normality test)

```{r}
shapiro.test(res)
```
>The residuals were uncorrelated but not normally distributed.
>The modification of model is needed.