Team 3 pull request

.Rbuildignore

@@ -0,0 +1,2 @@
+^.*\.Rproj$
+^\.Rproj\.user$

.gitignore

@@ -0,0 +1,4 @@
+.Rproj.user
+.Rhistory
+.RData
+.Ruserdata

DESCRIPTION

@@ -0,0 +1,11 @@
+Package: tree3
+Title: Package aimed to preprocess and build strong decision tree based predictions for deposit buying
+Version: 0.0.0.9000
+Authors@R: person("Group 3", "Data Scinece Academy", email = "first.last@example.com", role = c("aut", "cre"))
+Description: What the package does (one paragraph).
+Depends: R (>= 3.5.0)
+Imports: tidyverse, rpart, rpart.plot, DescTools
+License: GPL-2
+Encoding: UTF-8
+LazyData: true
+RoxygenNote: 6.1.1

NAMESPACE

@@ -0,0 +1,7 @@
+# Generated by roxygen2: do not edit by hand
+
+export(clean)
+export(model)
+export(model_accuracy)
+export(predictions)
+export(transform)

R/bmarketing.R

@@ -0,0 +1,31 @@
+#' Banking Marketing Data
+#'
+#' A dataset containing the prices and other attributes of almost 54,000
+#' diamonds.
+#'
+#' @format A data frame with 4119 rows and 21 variables:
+#' \describe{
+#'   \item{age}{age, integer}
+#'   \item{job}{factor}
+#'   \item{marital}{factor}
+#'   \item{education}{factor}
+#'   \item{default}{factor}
+#'   \item{housing}{factor}
+#'   \item{load}{factor}
+#'   \item{contact}{factor}
+#'   \item{month}{factor}
+#'   \item{day_of_week}{factor}
+#'   \item{duration}{integer}
+#'   \item{campaign}{integer}
+#'   \item{pdays}{integer}
+#'   \item{previous}{integer}
+#'   \item{poutcome}{factor}
+#'   \item{emp.var.rate}{numeric}
+#'   \item{cons.price.idx}{numeric}
+#'   \item{cons.conf.idx}{numeric}
+#'   \item{euribor3m}{numeric}
+#'   \item{nr.employed}{numeric}
+#'   \item{y}{target variable}
+#' }
+#' @source marketing
+"bmarketing"

R/clean.R

@@ -0,0 +1,43 @@
+#' Cleanin Data
+#'
+#' \code{clean} returns a clean data set.
+#'
+#' @param data A data.frame containing a target variable named y
+#' @param target_name A character containing the name of the target variable
+#' @return A cleaned data.frame. 
+#'
+#' This means:
+#' 
+#'   - return an error if the target variable contains any missing values (NA’s).
+#'   - Give clear warnings for all other variables which contain NA’s.
+#'   - Remove any columns (and report as warning) which contain more than 50% NA’s.
+#'   
+#' @examples
+#' data(bmarketing)
+#' clean(bmarketing,"y")
+#' @export
+
+clean<-function(data,target_name){
+
+  if(sum(is.na(data[,target_name]))>0) stop("Target variable y contains missing values (NA's)")
+
+  for(col in names(data)){
+
+    if(sum(is.na(data[,col]))>0){
+
+      if(sum(is.na(data[,col]))/nrow(data) > 0.5){
+
+        data <- data[,!(names(data)==col)]
+        warning(paste0("Variable ",col," removed due to more than 50% NA's"))
+
+      }else{
+        warning(paste0("Variable ",col," contains NA's"))
+      }
+
+    }
+
+  }
+
+  return(data)
+
+}

R/model.R

@@ -0,0 +1,25 @@
+#'  Runing decision tree model and ploting it results
+#'
+#' \code{clean} returns a decision tree model.
+#'
+#' @param input_data A data.frame containing a target variable named which name is defined as input argument
+#' @param target_name A character containing the name of the target variable
+#' @return A decision tree model 
+#'
+#' This means:
+#' 
+#'   Model: Create a decision tree model to predict if customer signing a term deposit.
+
+#' Model-Plot: We shall implement a function to present a nice representation of the model, e.g. for a decision tree we should plot the tree and respective nodes.
+#'   
+#' @examples
+#' data("bmarketing")
+#' dt_model <- model(input_data = bmarketing,target_name="y")
+#' @export
+
+model <- function(input_data,target_name) {
+  dt_model<- rpart::rpart(as.formula(paste(target_name," ~ .")), data = input_data)
+  rpart.plot::rpart.plot(dt_model)
+  return (dt_model)
+}
+

R/model_accuracy.R

@@ -0,0 +1,58 @@
+#' model_accuracy
+#'
+#' The function calculates accuracy (as a share of correctly classified observations and total observations), confusion matrix of a classification model
+#' , sensitivity and specificity. For further reference to confusion matrix, sensitivity and specificity please see \url{https://en.wikipedia.org/wiki/Confusion_matrix}
+#' , respective \url{https://en.wikipedia.org/wiki/Sensitivity_and_specificity}.
+#' 
+#' @param real a vector of realized observations 
+#' @param pred a vector of corresponding predictions
+#' @param chosenvar a string/numeric which choose for which class sensitivity will be calculated
+#' @return \code{accuracy} ... a number
+#' @return \code{confusion_matrix} ... a matrix
+#' @return \code{sensitivity} ... a number
+#' @return \code{specificity} ... a number
+#' @examples
+#' example_real=c(1,2,3,1,2,3,1,2,3) # a vector of realized observations
+#' example_pred=c(1,2,3,1,2,3,2,2,2) # a vector of predictions
+#' example_chosenvar=2
+#' model_accuracy(example_real,example_pred,example_chosenvar)
+#' @export
+
+model_accuracy <- function(real,pred,chosenvar){
+
+  #Tests if input is correct
+  if(length(which(!(unique(pred) %in% unique(real))))>0) stop('Predictions attain at least one value, which is not in realized data.')
+  if(length(real)!=length(pred)) stop('Different length of input vectors.')
+  if(any(is.na(real))==TRUE) stop('There is at least one NA in vector of realized data.')
+  if(any(is.na(pred))==TRUE) stop('There is at least one NA in vector of predicted data.')
+
+
+  ## Calculate the accuracy as a percentage of correctly predicted (classificated) observations.
+  acc_perc=mean(real == pred)
+
+  # Calculate confusion matrix.
+  confusion_matrix=table(real,pred)  
+
+  chosencol0<-colnames(confusion_matrix)==chosenvar
+  chosencol=which(chosencol0==TRUE)
+
+
+  #Calculate sensitivity and specificity
+  truepos<-sum(confusion_matrix[chosencol,chosencol])
+  falseneg<-sum(confusion_matrix[chosencol,-chosencol])
+  trueneg=sum(confusion_matrix[-chosencol,-chosencol])
+  falsepos=sum(confusion_matrix[-chosencol,chosencol])
+
+
+  sensitivity<-truepos/(truepos+falseneg)
+  specificity<-trueneg/(trueneg + falsepos)
+
+  #put results together
+  result<-list(acc_perc,confusion_matrix,sensitivity,specificity)
+  names(result) <-c('accuracy','confusion_matrix','sensitivity','specificity')
+
+  if(acc_perc<0.7) warning('Accuracy is below 70%.')
+
+  return(result)
+
+}

R/predictions.R

@@ -0,0 +1,16 @@
+#' Predictions.
+#'
+#' \code{predictions} returns predictions.
+#'
+#' @param dt_model model
+#' @param data data the model was generated with
+#' @examples
+#' data("bmarketing")
+#' dt_model <- model(input_data = bmarketing,target_name="y")
+#' predictions(dt_model, bmarketing)
+#' @export
+
+predictions <- function(dt_model, data)
+{
+  predict(dt_model, data, type = "class")
+}

R/testing.R

@@ -0,0 +1,17 @@
+# library(tree3)
+# 
+# data("bmarketing")
+# bmarketing
+# 
+# t2 <- clean(bmarketing)
+# t3 <- transform(t2)
+# 
+# dt_model <- model(t3)
+# 
+# pred <- predictions(dt_model,t3)
+# 
+# model_accuracy(t3$y,pred)
+# 
+# 
+# 
+# 

R/transform.R

@@ -0,0 +1,40 @@
+# Data Transformation
+#' Log transformation and factor transformation into numeric variables (and vice versa) as necessary.
+#' The function needs DescTools library installed
+#' 
+#' \code{transform} transforms the sent data to log (for numeric) or factor transformation into numeric, as necessary.
+#' 
+#' @param data A data.frame containing the target variable
+#' @param column The column, with double quotation marks which needs to be transformed
+#' @param option The option to replace NA's by the mean or by modus (1 - replace NA's - default, 0 - no replacement)
+#' @return The column from the data frame that is returned is the transformed data
+#'   
+#' @examples
+#' transform(bmarketing, "age", 1)
+#' @export
+
+transform <- function(data, column, option = 1) {
+
+  if(is.integer(data[,column]) || is.numeric(data[,column])) {
+  # Taking care of missing data & log
+    if (option == 1) {
+      Mean <- mean(data[,column], na.rm = TRUE)
+      y <- ifelse(is.na(data[,column]), Mean, data[,column])  
+    }
+    if (option == 0) {
+      y <- data[,column]
+    }
+
+    if(min(y, na.rm = TRUE) >=0)
+      y <- log(y)
+  }
+
+  # Encoding categorical data
+  if(is.factor(data[,column])) {
+    y <- as.numeric(data[,column])
+    if (option == 1) {
+      y <- ifelse(is.na(y), DescTools::Mode(y), y)
+    }
+  }
+  return(y)
+}

README.Rmd

@@ -17,7 +17,38 @@ knitr::opts_chunk$set(
 
 ## Overview
 
-The bmarketing dataset
+This package is dedicated to classification analysis using decision trees. Aside from decision tree model itself and prediction function it has got all necessary supporting tools - data cleaning, transformation function, plot of the results and calculation of accuracy of predictions and confusion matrix.
 
-<!-- TODO: Change README to make it more descriptive, add examples, etc. -->
+## Functionalities
 
+The functionalities are as follows.
+
+* Data Cleaning
+
+```{r}
+require("tree3")
+
+data("bmarketing")
+cleanedData <- clean(data = bmarketing,target_name = "y")
+```
+
+* Data Transforming
+```{r}
+cleanedData$cons.price.id <- transform(cleanedData,column = "cons.price.idx")
+```
+
+* Finding a Model
+```{r}
+treeModel <-  model(input_data = cleanedData,target_name="y")
+```
+
+
+* Getting the predictions
+```{r}
+predictionData <-  predictions(dt_model = treeModel,data = cleanedData)
+```
+
+* Assessing the model accuracy
+```{r}
+model_accuracy(real = cleanedData$y,pred = predictionData,chosenvar='yes')
+```

README.md

@@ -8,7 +8,79 @@ Status](https://img.shields.io/codecov/c/github/Quantargo/bmarketing/master.svg)
 
 ## Overview
 
-The bmarketing
-dataset
+This package is dedicated to classification analysis using decision
+trees. Aside from decision tree model itself and prediction function it
+has got all necessary supporting tools - data cleaning, transformation
+function, plot of the results and calculation of accuracy of predictions
+and confusion matrix.
 
-<!-- TODO: Change README to make it more descriptive, add examples, etc. -->
+## Functionalities
+
+The functionalities are as follows.
+
+  - Data Cleaning
+
+<!-- end list -->
+
+``` r
+require("tree3")
+#> Loading required package: tree3
+#> 
+#> Attaching package: 'tree3'
+#> The following object is masked from 'package:base':
+#> 
+#>     transform
+
+data("bmarketing")
+cleanedData <- clean(data = bmarketing,target_name = "y")
+```
+
+  - Data
+Transforming
+
+<!-- end list -->
+
+``` r
+cleanedData$cons.price.id <- transform(cleanedData,column = "cons.price.idx")
+```
+
+  - Finding a Model
+
+<!-- end list -->
+
+``` r
+treeModel <-  model(input_data = cleanedData,target_name="y")
+```
+
+![](man/figures/README-unnamed-chunk-4-1.png)<!-- -->
+
+  - Getting the predictions
+
+<!-- end list -->
+
+``` r
+predictionData <-  predictions(dt_model = treeModel,data = cleanedData)
+```
+
+  - Assessing the model
+accuracy
+
+<!-- end list -->
+
+``` r
+model_accuracy(real = cleanedData$y,pred = predictionData,chosenvar='yes')
+#> $accuracy
+#> [1] 0.9271668
+#> 
+#> $confusion_matrix
+#>      pred
+#> real    no  yes
+#>   no  3583   85
+#>   yes  215  236
+#> 
+#> $sensitivity
+#> [1] 0.5232816
+#> 
+#> $specificity
+#> [1] 0.9768266
+```