Assignment 3a

Author

Anastasiia Gmyrina

Approach

For this assignment I will use the movie ratings database from the previous assignment. The goal is to use Global Baseline Estimate to predict ratings for movies not yet seen by users. I will calculate the overall mean of all available ratings. Then I will calculate the average rating for each user and each movie and compare these averages to the global mean to find the user and movie effects. Then I will identify missing ratings, and assign them new predicted values.

Data loading

Show code

library(DBI)
library(RSQLite)
library(dplyr)
mydb <- dbConnect(SQLite(), "../week2/movie_ratings.db")
movie_ratings <- dbGetQuery(mydb, " 
SELECT users.name, movies.title, ratings.rating
FROM ratings
JOIN users 
    ON ratings.user_id = users.user_id
JOIN movies
    ON ratings.movie_id = movies.movie_id
ORDER BY users.user_id, movies.movie_id;")
movie_ratings

Global mean

I calculated the global mean using all available movie ratings while excluding missing values. This represents the overall average rating in the dataset.

Show code

global_mean <- movie_ratings %>%
  summarise(mean_rating = mean(rating, na.rm = TRUE)) %>%
  pull(mean_rating)
global_mean

[1] 3.96

User effect

I calculated the average rating for each user first and then user effect as the difference between each user’s average and the global mean, showing whether a user generally rates movies higher or lower than average.

Show code

user_avg <- movie_ratings %>%
  group_by(name) %>%
  summarise(user_avg = round(mean(rating, na.rm = TRUE),2),
            user_effect = user_avg - global_mean)
user_avg

Movie effect

Same strategy with movies

Show code

movie_avg <- movie_ratings %>%
  group_by(title) %>%
  summarise(movie_avg = round(mean(rating, na.rm = TRUE),2),
            movie_effect = movie_avg - global_mean)
movie_avg

Replacing missing values with global baseline estimate

Now I identified the missing ratings and used LEFT JOIN to combine them with the user and movie effects. I used the Global Baseline Estimate formula:

Predicted Rating = Global Mean + User Effect + Movie Effect

Show code

missing_ratings <- movie_ratings %>%
  filter(is.na(rating)) %>%
  left_join(user_avg, by = "name") %>%
  left_join(movie_avg, by = "title")
missing_ratings

Show code

missing_ratings %>%
  mutate(pred_rating = round(global_mean + user_effect + movie_effect, 2)) %>%
  select(name, title, pred_rating)

The Global Baseline Estimate was used to predict the five missing ratings based on the overall mean, user effect, and movie effect. The predicted ratings ranged from 3.19 to 5.21. One limitation of this method is that the predicted values are not restricted to the original 1–5 rating scale, which is why one prediction was higher than 5.

--- title: "Assignment 3a" author: "Anastasiia Gmyrina" editor: visual format: html: theme: cosmo toc: true toc-location: left toc-depth: 3 code-fold: true code-summary: "Show code" code-tools: true df-print: paged smooth-scroll: true embed-resources: true grid: body-width: 800px --- ## Approach For this assignment I will use the movie ratings database from the previous assignment. The goal is to use Global Baseline Estimate to predict ratings for movies not yet seen by users. I will calculate the overall mean of all available ratings. Then I will calculate the average rating for each user and each movie and compare these averages to the global mean to find the user and movie effects. Then I will identify missing ratings, and assign them new predicted values. ## Data loading ```{r} #| message: false library(DBI) library(RSQLite) library(dplyr) mydb <- dbConnect(SQLite(), "../week2/movie_ratings.db") movie_ratings <- dbGetQuery(mydb, " SELECT users.name, movies.title, ratings.rating FROM ratings JOIN users ON ratings.user_id = users.user_id JOIN movies ON ratings.movie_id = movies.movie_id ORDER BY users.user_id, movies.movie_id;") movie_ratings ``` ## Global mean I calculated the global mean using all available movie ratings while excluding missing values. This represents the overall average rating in the dataset. ```{r} global_mean <- movie_ratings %>% summarise(mean_rating = mean(rating, na.rm = TRUE)) %>% pull(mean_rating) global_mean ``` ## User effect I calculated the average rating for each user first and then user effect as the difference between each user's average and the global mean, showing whether a user generally rates movies higher or lower than average. ```{r} user_avg <- movie_ratings %>% group_by(name) %>% summarise(user_avg = round(mean(rating, na.rm = TRUE),2), user_effect = user_avg - global_mean) user_avg ``` ## Movie effect Same strategy with movies ```{r} movie_avg <- movie_ratings %>% group_by(title) %>% summarise(movie_avg = round(mean(rating, na.rm = TRUE),2), movie_effect = movie_avg - global_mean) movie_avg ``` ## Replacing missing values with global baseline estimate Now I identified the missing ratings and used LEFT JOIN to combine them with the user and movie effects. I used the Global Baseline Estimate formula: Predicted Rating = Global Mean + User Effect + Movie Effect ```{r} missing_ratings <- movie_ratings %>% filter(is.na(rating)) %>% left_join(user_avg, by = "name") %>% left_join(movie_avg, by = "title") missing_ratings ``` ```{r} missing_ratings %>% mutate(pred_rating = round(global_mean + user_effect + movie_effect, 2)) %>% select(name, title, pred_rating) ``` The Global Baseline Estimate was used to predict the five missing ratings based on the overall mean, user effect, and movie effect. The predicted ratings ranged from 3.19 to 5.21. One limitation of this method is that the predicted values are not restricted to the original 1–5 rating scale, which is why one prediction was higher than 5.