Recommendation Engines

Professor: Ivan Tarradellas Olmo

Done By: Hani Eid

Lab #2

Recommendation Sports

library(data.table)
library(xlsx)

## Loading required package: rJava

## Loading required package: xlsxjars

library(knitr)
library(kableExtra)

knitr::opts_knit$set(root.dir = 'C:/Users/Hani/Google Drive/IE/Term 2/5- Recommendation Engines/Assignments/2/Problem')
#setwd('C:/Users/Hani/Google Drive/IE/Term 2/5- Recommendation Engines/Assignments/2/Problem')

drawTable = function(DataFrame){
  DataFrame %>%
  kable("html") %>%
  kable_styling() %>%
  row_spec(row = 1:nrow(DataFrame), bold = T, color = "white", background = "#000000", align = "c") %>%
  column_spec(column = c(1,3), background = "#0AA19A") %>%
  row_spec(0, bold = T, align = "c", font_size = 18, color = "#B60404")  
}

createProfiles = function(sports, users) {
    profiles = sapply(sports, function(x) sapply(users, function(y) sum(x * y, na.rm = T)))
    return(profiles)
}

process = function(sports, profiles) {
    # Predictions
    sports_T = as.data.frame(t(sports))
    profiles_T = as.data.frame(t(profiles))
    numerator = sapply(profiles_T, function(x)(sapply(sports_T, function(y) sum(x * y))))
    profiles_Sqaured = profiles * profiles
    sports_Squared = sports * sports
    sports_Squared_Sum = rowSums(sports_Squared, na.rm = T)
    profiles_Sqaured_SUm = rowSums(profiles_Sqaured, na.rm = T)

    denom1 = sqrt(sports_Squared_Sum)
    denom2 = sqrt(profiles_Sqaured_SUm)
    denom_Final = sapply(denom2, function(x) sapply(denom1, function(y) sum(x * y, na.rm = T)))
    prediction_Table1 = as.data.frame(numerator / denom_Final)
    sum(numerator / denom_Final, na.rm = T)

    prediction_Table1[is.na(prediction_Table1)] = 0

    Likes1 = colSums(data.frame(Likes = sapply(prediction_Table1, function(x) x > 0)))
    Dislikes1 = colSums(data.frame(Dislikes = sapply(prediction_Table1, function(x) x < 0)))
    Neutral1 = colSums(data.frame(Neutral = sapply(prediction_Table1, function(x) x == 0)))

    newList <- list("Prediction" = prediction_Table1, "Likes" = Likes1, "Dislikes" = Dislikes1, "Neutral" = Neutral1)
    return(newList)
}

CB_Simply_Unary_Sports = read.xlsx("exerciseCB.xlsx", sheetName = "CB - Simply Unary", colIndex = 2:11, rowIndex = 2:22, header = T)
CB_Simply_Unary_Users = read.xlsx("exerciseCB.xlsx", sheetName = "CB - Simply Unary", colIndex = 15:18, rowIndex = 2:22, header = T)
CB_Simply_Unary_Predictions = read.xlsx("exerciseCB.xlsx", sheetName = "CB - Simply Unary", colIndex = 25:28, rowIndex = 2:22, header = T)

CB_Simply_Unary_Sports = as.data.frame(sapply(CB_Simply_Unary_Sports, as.numeric))


# Create User Profiles Table
profiles = createProfiles(CB_Simply_Unary_Sports, CB_Simply_Unary_Users)
sum(profiles, na.rm = T)

## [1] 2

# Predictions
output = process(CB_Simply_Unary_Sports, profiles)

Prediction = output$Prediction
Likes1 = output$Likes
Dislikes1 = output$Dislikes
Neutral = output$Neutral
rownames(Prediction) <- c()
drawTable(Prediction)

CB - Unit Weight:

CB_Unit_Weight = read.xlsx("exerciseCB.xlsx", sheetName = "CB - Unit Weight", colIndex = 2:11, rowIndex = 2:22, header = T)
CB_Unit_Weight_Users = read.xlsx("exerciseCB.xlsx", sheetName = "CB - Unit Weight", colIndex = 15:18, rowIndex = 2:22, header = T)

CB_Unit_Weight = CB_Unit_Weight / rowSums(CB_Unit_Weight, na.rm = T)
CB_Unit_Weight = as.data.frame(sapply(CB_Unit_Weight, as.numeric))

# Create User Profiles Table
profiles2 = createProfiles(CB_Unit_Weight, CB_Unit_Weight_Users)
sum(profiles2, na.rm = T)

## [1] -5.551115e-17

# Predictions
output2 = process(CB_Unit_Weight, profiles2)

Prediction2 = output2$Prediction
Likes2 = output2$Likes
Dislikes2 = output2$Dislikes
Neutral2 = output2$Neutral

rownames(Prediction2) <- c()

drawTable(Prediction2)

CB - IDF:

CB_IDF = read.xlsx("exerciseCB.xlsx", sheetName = "CB - IDF", colIndex = 2:11, rowIndex = 2:22, header = T)
CB_IDF_Users = read.xlsx("exerciseCB.xlsx", sheetName = "CB - IDF", colIndex = 15:18, rowIndex = 2:22, header = T)
CB_IDF_DF = colSums(CB_IDF, na.rm = T)
CB_IDF = CB_IDF / rowSums(CB_IDF)

IDF = log10(nrow(CB_IDF) / colSums(CB_IDF > 0))
CB_IDF = as.data.frame(sapply(CB_IDF, as.numeric))

# Create User Profiles Table
profiles3 = createProfiles(CB_IDF, CB_IDF_Users)
sum(profiles3, na.rm = T)

## [1] -5.551115e-17

profiles3 = as.data.frame(t(as.data.frame(t(profiles3)) * IDF))

# Predictions
output3 = process(CB_IDF, profiles3)

Prediction3 = output3$Prediction
Likes3 = output3$Likes
Dislikes3 = output3$Dislikes
Neutral3 = output3$Neutral

rownames(Prediction3) <- c()

drawTable(Prediction3)

Hybrid_Switching = read.xlsx("exerciseCB.xlsx", sheetName = "Hybrid - Switching", colIndex = 2:11, rowIndex = 2:22, header = T)

Hybrid_Challenge = read.xlsx("exerciseCB.xlsx", sheetName = "Hybrid - Challenge", colIndex = 2:11, rowIndex = 2:22, header = T)