RamishQ3(Updated).Rmd

---
title: "Q3"
author: "Ramish"
date: "20/09/2019"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(plyr)
library(dplyr)
library(arules)
```

```{r}
c <- read.csv("Practice/Data/courses.csv", header = TRUE, sep=",", stringsAsFactors = TRUE)
a <- read.csv("Practice/Data/assessments.csv", header = TRUE, sep=",", stringsAsFactors = TRUE)
s <- read.csv("Practice/Data/studentAssessment.csv", header = TRUE, sep=",", stringsAsFactors = TRUE)

```

```{r}

m1 <- merge(c, a, by = c("code_module", "code_presentation")) %>% merge(s, by = "id_assessment")
```

```{r}

t1 <- m1 %>% group_by(code_module, code_presentation, weight, score)

t1$date_submitted <- NULL
t1$is_banked <- NULL
t1$module_presentation_length <- NULL
t1$date <- NULL
t1 <- t1[complete.cases(t1), ]


t1$code_module <- as.factor(t1$code_module)
t1$code_presentation <- as.factor(t1$code_presentation)
t1$weight <- as.factor(t1$weight)
t1$score <- as.factor(t1$score)
t1

```

```{r}

rules <- apriori(t1, parameter = list(support = 0.01, confidence = 0.15))


subRules <- which(colSums(is.subset(rules, rules)) > 1) # get subset rules in a vector
length(subRules)
rules <- rules[-subRules] # remove subset rules

rules_sup <- sort (rules, by="support", decreasing = TRUE) # sort by support
inspect(head(rules_sup)) # print top 5 rules

```

The datasets used for this analysis were courses, assessments and studentAssessments. The methodology was to merge two sets first and then merge the result with last set. Moving forward, the merged rows had redundancy in them which had to be removed and only the rows which helped in studying the data needed to emerge. hence, a table t1 was created which deleted the unnecessary rows of data like date_submitted, module_presentation_length etc. The remaining rows were formatted as factors which would later on help for association rules. The apriori algorithm was applied to form rules on the derived table set from eariler on and study the support and confidence applied on them. Redundant data was again a problem but was solved by creating a subset (subRules) which removed the repeating values. Finally, the last step was to sort our rules with support in descending order. 
The result was rather disappointing as no finding was obtained from the rules generated.