Assignment 3
2024-11-26
Set Working Directory
getwd()## [1] "/Users/markyuhasz/FallWinter 2024 Classes/MKTG3P98-Business analytics and intell/Assignment 3"setwd("/Users/markyuhasz/FallWinter 2024 Classes/MKTG3P98-Business analytics and intell/Assignment 3")Activate Libraries that are Going to be Used
library(readxl)
library(ggplot2)
library(readr) # To Read CSV Files
library(stringr)
library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionlibrary(janitor)##
## Attaching package: 'janitor'## The following objects are masked from 'package:stats':
##
## chisq.test, fisher.testlibrary(performance)
library(patchwork)
library(ggplotify)
library(reshape2)
library(tidyr)##
## Attaching package: 'tidyr'## The following object is masked from 'package:reshape2':
##
## smithsImport the Seperate Data Sets
Movie_General<-read.csv("Movie Dataset_General Audience.csv")
Movie_Financial<-read.csv("Movie Dataset_Financials.csv")Viewing the Data Sets
View(Movie_General)
View(Movie_Financial)Merge the Data Sets
Movie_Total<-merge(Movie_Financial,Movie_General,by= "original_title")Find Any Duplicates in Original Title Name
Movie_Total_Clean<-clean_names(Movie_Total)
Movie_Total_Clean%>% get_dupes(original_title)## original_title dupe_count budget_millions
## 1 Fantastic Four\xca 4 17.0
## 2 Fantastic Four\xca 4 17.0
## 3 Fantastic Four\xca 4 13.0
## 4 Fantastic Four\xca 4 13.0
## 5 Godzilla Resurgence\xca 4 8.9
## 6 Godzilla Resurgence\xca 4 8.9
## 7 Godzilla Resurgence\xca 4 1.9
## 8 Godzilla Resurgence\xca 4 1.9
## 9 Hercules\xca 4 20.0
## 10 Hercules\xca 4 20.0
## 11 Hercules\xca 4 6.0
## 12 Hercules\xca 4 6.0
## 13 Pan\xca 4 4.0
## 14 Pan\xca 4 4.0
## 15 Pan\xca 4 3.0
## 16 Pan\xca 4 3.0
## 17 The Fast and the Furious\xca 4 35.0
## 18 The Fast and the Furious\xca 4 35.0
## 19 The Fast and the Furious\xca 4 20.0
## 20 The Fast and the Furious\xca 4 20.0
## 21 The Legend of Tarzan\xca 4 35.0
## 22 The Legend of Tarzan\xca 4 35.0
## 23 The Legend of Tarzan\xca 4 6.0
## 24 The Legend of Tarzan\xca 4 6.0
## 25 The Twilight Saga: Breaking Dawn - Part 2\xca 4 1.0
## 26 The Twilight Saga: Breaking Dawn - Part 2\xca 4 1.0
## 27 The Twilight Saga: Breaking Dawn - Part 2\xca 4 0.5
## 28 The Twilight Saga: Breaking Dawn - Part 2\xca 4 0.5
## revenue_millions language country type genre runtime
## 1 136.62 English USA Feature Film Mystery & Suspense 106
## 2 136.62 English USA Feature Film Drama 101
## 3 45.30 English USA Feature Film Mystery & Suspense 106
## 4 45.30 English USA Feature Film Drama 101
## 5 4.16 Japanese Japan Feature Film Comedy 93
## 6 4.16 Japanese Japan Feature Film Drama 98
## 7 157.11 Japanese Japan Feature Film Comedy 93
## 8 157.11 Japanese Japan Feature Film Drama 98
## 9 34.08 English USA Feature Film Drama 105
## 10 34.08 English USA Documentary Documentary 107
## 11 1.27 English USA Feature Film Drama 105
## 12 1.27 English USA Documentary Documentary 107
## 13 1.60 English USA Feature Film Drama 130
## 14 1.60 English USA Feature Film Other 90
## 15 0.70 English USA Feature Film Drama 130
## 16 0.70 English USA Feature Film Other 90
## 17 77.48 English USA Feature Film Drama 104
## 18 77.48 English USA Feature Film Drama 95
## 19 18.22 English USA Feature Film Drama 104
## 20 18.22 English USA Feature Film Drama 95
## 21 82.35 English USA Documentary Documentary 92
## 22 82.35 English USA Feature Film Drama 94
## 23 11.12 English USA Documentary Documentary 92
## 24 11.12 English USA Feature Film Drama 94
## 25 3.39 English USA Documentary Documentary 98
## 26 3.39 English USA Feature Film Comedy 95
## 27 1025.47 English USA Documentary Documentary 98
## 28 1025.47 English USA Feature Film Comedy 95
## mpaa_rating imdb_rating imdb_num_votes critics_rating critics_score
## 1 R 6.3 25054 Rotten 43
## 2 R 2.1 9904 Fresh 35
## 3 R 6.3 25054 Rotten 43
## 4 R 2.1 9904 Fresh 35
## 5 PG 4.2 182983 Certified Fresh 34
## 6 PG-13 6.3 50340 Fresh 60
## 7 PG 4.2 182983 Certified Fresh 34
## 8 PG-13 6.3 50340 Fresh 60
## 9 R 3.0 9216 Rotten 50
## 10 Unrated 3.8 10522 Fresh 21
## 11 R 3.0 9216 Rotten 50
## 12 Unrated 3.8 10522 Fresh 21
## 13 R 2.0 9216 Rotten 34
## 14 PG 3.6 1010 Fresh 33
## 15 R 2.0 9216 Rotten 34
## 16 PG 3.6 1010 Fresh 33
## 17 R 7.1 128361 Fresh 68
## 18 R 7.2 35635 Certified Fresh 93
## 19 R 7.1 128361 Fresh 68
## 20 R 7.2 35635 Certified Fresh 93
## 21 Unrated 6.8 1942 Certified Fresh 66
## 22 R 3.4 57933 Rotten 25
## 23 Unrated 6.8 1942 Certified Fresh 66
## 24 R 3.4 57933 Rotten 25
## 25 Unrated 3.1 1010 Fresh 45
## 26 PG 7.5 880 Fresh 90
## 27 Unrated 3.1 1010 Fresh 45
## 28 PG 7.5 880 Fresh 90
## audience_rating audience_score best_pic_nom facebook_likes
## 1 Spilled 49 no 1261
## 2 Upright 26 no 51261
## 3 Spilled 49 no 1261
## 4 Upright 26 no 51261
## 5 Upright 33 yes 5699
## 6 Spilled 67 no 699
## 7 Upright 33 yes 5699
## 8 Spilled 67 no 699
## 9 Upright 40 no 16235
## 10 Upright 20 no 235
## 11 Upright 40 no 16235
## 12 Upright 20 no 235
## 13 Upright 24 no 393
## 14 Upright 24 no 393
## 15 Upright 24 no 393
## 16 Upright 24 no 393
## 17 Upright 78 no 125327
## 18 Upright 80 no 45327
## 19 Upright 78 no 125327
## 20 Upright 80 no 45327
## 21 Upright 68 no 121175
## 22 Spilled 39 no 11175
## 23 Upright 68 no 121175
## 24 Spilled 39 no 11175
## 25 Upright 46 no 1177
## 26 Upright 89 no 359177
## 27 Upright 46 no 1177
## 28 Upright 89 no 359177Delete all Duplicate Movie Titles
Movie_Total_Unique <- Movie_Total_Clean %>% distinct(original_title, .keep_all = TRUE)
print(head(Movie_Total_Unique, 10))## original_title budget_millions revenue_millions language
## 1 10,000 B.C.\xca 12.6 18.66
## 2 102 Dalmatians\xca 45.0 60.22 English
## 3 2 Fast 2 Furious\xca 16.0 31.56 English
## 4 2012\xca 3.5 0.75 English
## 5 300: Rise of an Empire\xca 16.0 19.68 English
## 6 47 Ronin\xca 50.0 240.36 English
## 7 50 First Dates\xca 50.0 36.35 English
## 8 A Beautiful Mind\xca 26.0 43.32 English
## 9 A Christmas Carol\xca 65.0 235.67 English
## 10 A Good Day to Die Hard\xca 8.5 5.77 English
## country type genre runtime mpaa_rating imdb_rating
## 1 Feature Film Drama 134 R 6.8
## 2 USA Feature Film Drama 108 PG 4.9
## 3 USA Feature Film Mystery & Suspense 97 PG-13 6.3
## 4 USA Feature Film Comedy 98 PG-13 6.3
## 5 USA Feature Film Action & Adventure 111 PG-13 6.0
## 6 USA Feature Film Drama 106 PG 7.8
## 7 USA Feature Film Comedy 87 R 5.4
## 8 USA Feature Film Action & Adventure 83 G 7.6
## 9 USA Feature Film Drama 100 R 7.0
## 10 USA Feature Film Action & Adventure 86 R 4.1
## imdb_num_votes critics_rating critics_score audience_rating audience_score
## 1 9025 Fresh 60 Upright 76
## 2 5136 Rotten 5 Spilled 13
## 3 54771 Rotten 40 Spilled 49
## 4 8646 Certified Fresh 44 Spilled 54
## 5 103789 Rotten 51 Spilled 51
## 6 12450 Fresh 75 Spilled 85
## 7 6811 Rotten 35 Spilled 31
## 8 78862 Rotten 50 Upright 81
## 9 8320 Rotten 70 Spilled 74
## 10 739 Fresh 53 Upright 42
## best_pic_nom facebook_likes
## 1 no 23343
## 2 no 84182
## 3 no 35296
## 4 no 445
## 5 no 21583
## 6 no 20965
## 7 no 12952
## 8 no 52827
## 9 no 48878
## 10 no 5481Catagorize and Group Based on Genre
Movie_Total_Unique <- Movie_Total_Unique %>% #Call Dataframe and Create New Column with New Groupings
mutate(Genre1 = case_when(genre == "Science Fiction & Fantasy" ~ "SciFi",
genre == "Mystery & Suspense" ~ "Mystery",
genre == "Drama" ~ "Drama",
genre == "Documentary" ~ "Documentary",
genre == "Comedy" ~ "Comedy",
genre == "Art House & International" ~ "Arts",
genre == "Other" ~ "Other",
genre == "Action & Adventure" ~ "Action",
genre == "Animation" ~ "Animation",
genre == "Horror" ~ "Horror",
genre == "Musical & Performing Arts" ~ "Arts",
TRUE ~ "Check"))Mutate Dataset for Predictive Modeling
MT <- Movie_Total_Unique %>%
mutate(
budget_log = log(budget_millions),
revenue_log = log(revenue_millions)
)Transform IMDB rating and correct outliers to be out of 100 so same scale is used
#Changes the scale of the imdb rating to a scale of 100
MT$imdb_rating <- MT$imdb_rating * 10
MT$genre <- as.factor(MT$genre)
# replace outliers of IMDB score that are greater than 100
MT$imdb_rating[MT$imdb_rating > 100]## [1] 550 440# which rows are responsible for the over 100
outlier <- MT[MT$imdb_rating>100,]
print(outlier)## original_title budget_millions revenue_millions language
## 447 The Amazing Spider-Man\xca 30 91.71 English
## 638 X-Men: The Last Stand\xca 6 35.40 English
## country type genre runtime mpaa_rating imdb_rating
## 447 USA Documentary Documentary 82 Unrated 550
## 638 Canada Feature Film Documentary 112 R 440
## imdb_num_votes critics_rating critics_score audience_rating audience_score
## 447 3459 Fresh 55 Upright 56
## 638 4908 Rotten 40 Spilled 47
## best_pic_nom facebook_likes Genre1 budget_log revenue_log
## 447 no 2489 Documentary 3.401197 4.518631
## 638 no 21714 Documentary 1.791759 3.566712# change the outliers numbers to the actual numbers based on the website
MT$imdb_rating[c(447, 638)] <- c(69, 66)Rename the revenue column to a more manageable name
Movie_Total <- clean_names(Movie_Total)
colnames(Movie_Total)## [1] "original_title" "budget_millions" "revenue_millions" "language"
## [5] "country" "type" "genre" "runtime"
## [9] "mpaa_rating" "imdb_rating" "imdb_num_votes" "critics_rating"
## [13] "critics_score" "audience_rating" "audience_score" "best_pic_nom"
## [17] "facebook_likes"Movie_Total <- Movie_Total %>%
rename(revenue_millions = revenue_millions)Renaming the column revenue_millions for better readability and consistency.
We want to answer how are certain factors driving overall revenue (critics score, audience, runtime, imdb and genre)
Movie_Data <- Movie_Total %>%
select(revenue_millions, critics_score, audience_score, runtime, imdb_rating, genre)Extracting only the columns relevant to the analysis to streamline further operations.
Remove NA values to prevent errors during analysis
Movie_Data <- na.omit(Movie_Data)Analyze average factors by genre category
Factors_by_Genre <- Movie_Data %>%
group_by(genre) %>%
summarise(
Avg_Critics_Score = mean(critics_score, na.rm = TRUE),
Avg_Audience_Score = mean(audience_score, na.rm = TRUE),
Avg_Runtime = mean(runtime, na.rm = TRUE),
Avg_IMDb_Rating = mean(imdb_rating, na.rm = TRUE),
Avg_Revenue = mean(revenue_millions, na.rm = TRUE)
)Calculating the average values of key factors for each genre to identify general trends across genres.
Visualize average revenue by genre
ggplot(Factors_by_Genre, aes(x = genre, y = Avg_Revenue, fill = genre)) +
geom_bar(stat = "identity") +
labs(title = "Average Revenue by Genre", x = "Genre", y = "Average Revenue (Millions)") +
theme_minimal()
### Interpretation: #### Animation leads as the highest
revenue-generating genre, followed by Science Fiction & Fantasy and
Mystery & Suspense, showcasing their strong audience appeal.
Conversely, Art House & International and Documentaries have lower
revenues, reflecting their niche markets. Most other genres, like Comedy
and Drama, exhibit stable mid-range performance, suggesting balanced but
less exceptional financial returns.
Creating a scatter plot to explore the relationship between critics’ scores and revenue for each genre.
ggplot(Movie_Data, aes(x = critics_score, y = revenue_millions, color = genre)) +
geom_point(alpha = 0.7) +
labs(title = "Critics Score vs Revenue by Genre", x = "Critics Score", y = "Revenue (Millions)") +
theme_minimal()
Identify correlations between factors and revenue for each genre
Top_Factors_By_Genre <- Movie_Data %>%
group_by(genre) %>%
summarise(
Cor_Critics = cor(revenue_millions, critics_score, use = "complete.obs"),
Cor_Audience = cor(revenue_millions, audience_score, use = "complete.obs"),
Cor_Runtime = cor(revenue_millions, runtime, use = "complete.obs"),
Cor_IMDb = cor(revenue_millions, imdb_rating, use = "complete.obs")
) %>%
pivot_longer(cols = starts_with("Cor"), names_to = "Factor", values_to = "Correlation") %>%
arrange(genre, desc(Correlation))Calculating and reshaping correlations to understand how each factor affects revenue by genre.
Print top factors by genre
print(Top_Factors_By_Genre)## # A tibble: 44 × 3
## genre Factor Correlation
## <chr> <chr> <dbl>
## 1 Action & Adventure Cor_Audience 0.466
## 2 Action & Adventure Cor_IMDb 0.464
## 3 Action & Adventure Cor_Runtime 0.354
## 4 Action & Adventure Cor_Critics 0.337
## 5 Animation Cor_Runtime 0.772
## 6 Animation Cor_Critics 0.525
## 7 Animation Cor_Audience 0.431
## 8 Animation Cor_IMDb 0.415
## 9 Art House & International Cor_Runtime 0.751
## 10 Art House & International Cor_Audience 0.421
## # ℹ 34 more rowsDisplaying the full table of correlations to identify the strongest factors for each genre.
interpretation:
1. Audience Scores Drive Revenue Across All Genres, consistently show strong correlations with revenue, particularly in Science Fiction & Fantasy (0.796), Mystery & Suspense (0.575), and Comedy (0.559). This highlights that audience satisfaction is the most critical driver of financial success.
2. Runtime Impacts Certain Genres Differently:
Positive Impact: Longer runtimes are highly beneficial for Animation, Art House & International and Science Fiction & Fantasy
Negative Impact: Shorter runtimes are preferred in Horror and Musical & Performing Arts
3. Critics’ Scores Are Less Influential Than Audience Scores
While critics’ scores do impact revenue, their influence is typically weaker compared to audience feedback. For example, in Horror, Critics’ Scores (0.477) are less impactful than IMDb Ratings (0.466) and Audience Scores (0.450).
Plot top two driving factors for each genre
ggplot(Top_Factors_By_Genre, aes(x = genre, y = Correlation, fill = Factor)) +
geom_bar(stat = "identity", position = "dodge") +
labs(title = "Top Factors Driving Revenue by Genre", x = "Genre", y = "Correlation") +
theme_minimal()
#### To maximize box office revenue: Prioritize audience engagement
through targeted marketing and viewer-centric storytelling. Tailor
runtimes by genre, investing in longer narratives for Animation and
Sci-Fi, while keeping Horror and Musicals short. Leverage IMDb Ratings
as a quality benchmark to ensure broad public appeal.
#Question 1: Do Some Genres Generate More Revenue Than Others ## Create Table
Avg_Revenue <- Movie_Total_Unique %>%
group_by(Genre1) %>%
summarise(mean_revenue = mean(revenue_millions, na.rm = TRUE))
print(Avg_Revenue)## # A tibble: 10 × 2
## Genre1 mean_revenue
## <chr> <dbl>
## 1 Action 129.
## 2 Animation 237.
## 3 Arts 132.
## 4 Comedy 128.
## 5 Documentary 120.
## 6 Drama 140.
## 7 Horror 189.
## 8 Mystery 133.
## 9 Other 180.
## 10 SciFi 67.2Graph
ggplot(Avg_Revenue, aes(x = Genre1, y = mean_revenue)) +
geom_bar(stat = "identity") +
theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
labs(title = "Average Revenue by Genre", x = "Genre", y = "Average Revenue (Millions)")
### Therefore, Animation films tend to generate the most average
revenue.
Question 2: Do Critics have a Particular Preference for Certain Genres?
Create Table
Avg_Critic_Score <- Movie_Total_Unique %>%
group_by(Genre1) %>%
summarise(mean_critics_score = mean(critics_score, na.rm = TRUE))Graph
ggplot(Avg_Critic_Score, aes(x = Genre1, y = mean_critics_score)) +
geom_bar(stat = "identity") +
theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
labs(title = "Average Critics Score by Genre", x = "Genre", y = "Average Critics Score")
### Therefore, Critics tend to have a preference for Documentary
Films
Question 3: Is Critic Score more Important than Viewer Score?
Correlation Analysis
cor_critics <- cor(Movie_Total_Unique$critics_score, Movie_Total_Unique$revenue_millions, use = "complete.obs")cor_audience <- cor(Movie_Total_Unique$audience_score, Movie_Total_Unique$revenue_millions, use = "complete.obs")cat("Correlation with Critics Score:", cor_critics, "\n")## Correlation with Critics Score: 0.392774cat("Correlation with Audience Score:", cor_audience, "\n")## Correlation with Audience Score: 0.4428168Therefore, Audience score has a stronger linear relationshio with revenue (millions)
Multiple Linear Regression
lm_model <- lm(revenue_millions ~ critics_score + audience_score, data = Movie_Total_Unique)
summary(lm_model)##
## Call:
## lm(formula = revenue_millions ~ critics_score + audience_score,
## data = Movie_Total_Unique)
##
## Residuals:
## Min 1Q Median 3Q Max
## -279.37 -88.33 -25.64 29.13 1835.76
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -119.6096 21.4587 -5.574 3.67e-08 ***
## critics_score 1.1282 0.5019 2.248 0.0249 *
## audience_score 3.4444 0.5540 6.218 9.10e-10 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 186.3 on 641 degrees of freedom
## Multiple R-squared: 0.2024, Adjusted R-squared: 0.1999
## F-statistic: 81.32 on 2 and 641 DF, p-value: < 2.2e-16Critics Score (1.1282): For every 1- unit increase in critics score, the revenue is expected to increase by 1.13 million dollars. The p-value (0.0249) is less than 0.05 meaning it is statistically significant
Audience Score (3.4444): For every 1- unit increase in audience score, the revenue is expected to increase by 3.44 million dollars. The p-value (9.10e - 10) is less than 0.05 meaning it is highly statistically significant
Audience score has a stronger impact on revenue compared to Critics score
Interaction Effects
# Add an interaction term
lm_interaction <- lm(revenue_millions ~ critics_score * audience_score, data = Movie_Total_Unique)
summary(lm_interaction)##
## Call:
## lm(formula = revenue_millions ~ critics_score * audience_score,
## data = Movie_Total_Unique)
##
## Residuals:
## Min 1Q Median 3Q Max
## -299.02 -84.21 -25.48 20.35 1834.23
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -59.2411 51.9350 -1.141 0.2544
## critics_score -0.1927 1.1501 -0.168 0.8670
## audience_score 2.3249 1.0373 2.241 0.0254 *
## critics_score:audience_score 0.0217 0.0170 1.276 0.2023
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 186.2 on 640 degrees of freedom
## Multiple R-squared: 0.2044, Adjusted R-squared: 0.2007
## F-statistic: 54.81 on 3 and 640 DF, p-value: < 2.2e-16The p- value (0.2023) is greater than 0.05 menaing the interaction between Critics score and Audience score is not statistically significant, meaning their combimed influeunce does not significally impact revneue beyond their individual effects
Bar Plot of Coefficients
#Coefficients from the Regression Model
coefficients <- data.frame(
Variable = c("Critics Score", "Audience Score"),
Coefficient = c(1.1282, 3.4444)
)# Coefficients from the regression model
coefficients <- data.frame(
Variable = c("Critics Score", "Audience Score"),
Coefficient = c(1.1282, 3.4444)
)
# Bar plot
library(ggplot2)
ggplot(coefficients, aes(x = Variable, y = Coefficient, fill = Variable)) +
geom_bar(stat = "identity", width = 0.6) +
labs(
title = "Impact of Critics and Audience Scores on Revenue",
y = "Coefficient Value",
x = "Variable"
) +
theme_minimal() +
scale_fill_manual(values = c("Critics Score" = "pink", "Audience Score" = "blue")) +
geom_text(aes(label = round(Coefficient, 2)), vjust = -0.2)
### Audience Score is more important than Critics Score for predicting
film revenue. It has a larger coefficient, meaning its influence on
revenue is approximately 3 times greater than that of Critics Score.
Graph Audience Score vs Critics Score in Relation to Genre
options(repr.plot.res = 200, repr.plot.height = 5, repr.plot.width = 8)
ggplot(Movie_Total_Unique, aes(x = critics_score, y = audience_score)) +
geom_point(aes(colour = genre), alpha = 0.7) +
geom_smooth(method = "lm", se = FALSE, colour = "black") +
labs(
title = " Audience's Score vs Critics' Score",
x = "Critics Score",
y = "Audience Score",
colour = "Genre"
)## `geom_smooth()` using formula = 'y ~ x'
### Movies with higher critics’ scores tend to have higher audience
scores and vice versa, as seen from the upward-sloping trendline.
Visualize Correlation Between Audience Score and Critics Score and Revenue
correlation_matrix <- cor(Movie_Total_Unique[, c("critics_score", "audience_score", "revenue_millions")], use = "complete.obs")
print(correlation_matrix)## critics_score audience_score revenue_millions
## critics_score 1.0000000 0.7734838 0.3927740
## audience_score 0.7734838 1.0000000 0.4428168
## revenue_millions 0.3927740 0.4428168 1.0000000Plot Correlation Matrix
library(reshape2)
melted_correlation <- melt(correlation_matrix)
ggplot(melted_correlation, aes(x = Var1, y = Var2, fill = value)) +
geom_tile() +
scale_fill_gradient2(low = "blue", high = "red", mid = "white", midpoint = 0, limit = c(-1, 1)) +
labs(title = "Correlation Matrix", x = "", y = "") +
theme_minimal()
### There is a strong positive correlation between Critics’ score and
Audience score, as indicated by the darker red ### The correlation
between revenue and Audience score is stronger (darker red) than the
correlation between revenue and Critics’ score. ### While Critics’ score
shows a positive correlation with revenue, the relationship appears
weaker compared to Audience score, as suggested by the lighter shade of
red.
Graph Each Predictor
# Critics' Score vs Revenue
ggplot(Movie_Total_Unique, aes(x = critics_score, y = log10(revenue_millions))) +
geom_point(alpha = 0.5) +
geom_smooth(method = "lm", colour = "red") +
labs(title = "Critics' Score vs Revenue (Log Scale)", x = "Critics Score", y = "Log10 Revenue")## `geom_smooth()` using formula = 'y ~ x'
The graph shows a weak to moderate positive correlation between Critics’ Score and Revenue, indicating that higher Critics’ scores slightly increase movie revenue, but other factors likely play a more significant role.
# Audience Score vs Revenue
ggplot(Movie_Total_Unique, aes(x = audience_score, y = log10(revenue_millions))) +
geom_point(alpha = 0.5) +
geom_smooth(method = "lm", colour = "blue") +
labs(title = "Audience Score vs Revenue (Log Scale)", x = "Audience Score", y = "Log10 Revenue")## `geom_smooth()` using formula = 'y ~ x'
Moderate positive correlation between Audience Score and Revenue, indicating that higher Audience scores are more strongly associated with increased movie revenue compared to Critics’ scores
Comparing R- Squared Values for Individual Predictors
model_critics <- lm(revenue_millions ~ critics_score, data = Movie_Total_Unique)
model_audience <- lm(revenue_millions ~ audience_score, data = Movie_Total_Unique)
summary_critics <- summary(model_critics)
summary_audience <- summary(model_audience)
cat("R-squared for Critics Score Model:", summary_critics$r.squared, "\n")## R-squared for Critics Score Model: 0.1542714cat("R-squared for Audience Score Model:", summary_audience$r.squared, "\n")## R-squared for Audience Score Model: 0.1960867About 15.4% of the variance in movie revenue is explained by Critics’ scores alone.
About 19.6% of the variance in movie revenue is explained by Audience scores alone.
Neither variable explains the majority of the variance in revenue (both R-squared values are below 0.2)
QUESTION 4: Audience Score vs Critics’ Score by Movie Budget
Find Max and Min Budget Value
max_budget <- max(Movie_Total_Unique$budget_millions, na.rm = TRUE) # Remove NA values
min_budget <- min(Movie_Total_Unique$budget_millions, na.rm = TRUE)
cat("Maximum budget:", max_budget, "\n")## Maximum budget: 280cat("Minimum budget:", min_budget, "\n")## Minimum budget: 0.32Break Budget Down into 4 Catagories
# Define 4 range groups
Movie_Total_Unique$budget_category <- cut(
Movie_Total_Unique$budget_millions,
breaks = c(0.32, 70, 140, 210, 280),
labels = c("Low", "Moderate", "High", "Very High"),
include.lowest = TRUE
)Graph
options(repr.plot.res = 100, repr.plot.height = 5, repr.plot.width = 8)
ggplot(Movie_Total_Unique, aes(x = critics_score, y = audience_score)) +
geom_point(aes(colour = budget_category)) +
facet_wrap(~genre) +
labs(
title = "Audience Score vs Critics' Score by Budget Category",
x = "Critics' Score",
y = "Audience Score",
colour = "Budget Category"
)
The connection between Critics’ Scores and Audience Scores changes depending on the genre. Higher-budget genres like Drama and Action & Adventure tend to have more agreement, while lower-budget genres like Horror and Sci-Fi often show bigger differences. Documentaries have the most overlap between what critics and audiences like.
Look at Log regression compared to unmodified regression
ggplot(MT, aes(x = budget_millions, y = revenue_millions)) +
geom_point() +
geom_smooth(method = 'lm', se = FALSE, color = 'red')## `geom_smooth()` using formula = 'y ~ x'
ggplot(MT, aes(x = budget_log, y = revenue_log)) +
geom_point() +
geom_smooth(method = 'lm', se = FALSE, color = 'red')## `geom_smooth()` using formula = 'y ~ x'
Visualize Relationship of variables as they relate to Revenue
ggplot(MT) + aes(y = revenue_log, x = runtime, colour = audience_score, size = critics_score) +
geom_point() +
scale_color_gradientn(colors = c("red", "yellow", "green")) +
labs(
y = "Generated Revenue (Millions)",
x = "Runtime (Min)",
color = "Audience Score",
size = "Critic Score"
) +
theme_minimal() +
facet_wrap(~ genre)## Warning: Removed 1 row containing missing values or values outside the scale range
## (`geom_point()`).
Another Visualization of Variables as they relate to Revenue
ggplot(MT) + aes(y = revenue_log, x = budget_log, colour = imdb_rating, size = runtime) +
geom_point() +
scale_color_gradientn(colors = c("red", "yellow", "green")) +
labs(
y = "Generated Revenue (Millions)",
x = "Budget (Millions)",
color = "IMDB Rating (1-100)",
size = "Runtime (Min)"
) +
theme_minimal() +
facet_wrap(~ genre)## Warning: Removed 1 row containing missing values or values outside the scale range
## (`geom_point()`).
## Develop Multiple Linear Regression to see Pr(>|t|) ## Multiple
Linear regression of all variables
MTM1<- lm(revenue_log ~ budget_log+facebook_likes+critics_score+audience_score+imdb_rating+runtime+genre+mpaa_rating+imdb_num_votes, data = MT)
summary(MTM1)##
## Call:
## lm(formula = revenue_log ~ budget_log + facebook_likes + critics_score +
## audience_score + imdb_rating + runtime + genre + mpaa_rating +
## imdb_num_votes, data = MT)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.1261 -0.5296 0.0487 0.5997 3.1415
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.073e-01 3.616e-01 0.573 0.56659
## budget_log 4.415e-01 3.542e-02 12.465 < 2e-16 ***
## facebook_likes 5.001e-06 4.621e-07 10.822 < 2e-16 ***
## critics_score 8.549e-03 2.726e-03 3.136 0.00179 **
## audience_score 2.980e-03 4.160e-03 0.716 0.47411
## imdb_rating 2.924e-02 4.378e-03 6.679 5.35e-11 ***
## runtime -2.178e-03 2.208e-03 -0.986 0.32443
## genreAnimation 1.841e-02 3.840e-01 0.048 0.96178
## genreArt House & International -6.032e-02 2.983e-01 -0.202 0.83983
## genreComedy -9.428e-04 1.640e-01 -0.006 0.99542
## genreDocumentary -3.953e-01 2.172e-01 -1.820 0.06926 .
## genreDrama -2.658e-01 1.401e-01 -1.898 0.05817 .
## genreHorror -1.507e-01 2.455e-01 -0.614 0.53963
## genreMusical & Performing Arts -4.941e-01 3.131e-01 -1.578 0.11506
## genreMystery & Suspense -1.871e-01 1.833e-01 -1.021 0.30785
## genreOther -4.603e-02 2.929e-01 -0.157 0.87520
## genreScience Fiction & Fantasy -1.782e-01 3.518e-01 -0.507 0.61260
## mpaa_ratingNC-17 7.878e-01 7.451e-01 1.057 0.29074
## mpaa_ratingPG 1.860e-01 2.701e-01 0.689 0.49120
## mpaa_ratingPG-13 3.529e-01 2.762e-01 1.278 0.20184
## mpaa_ratingR 2.308e-01 2.673e-01 0.864 0.38816
## mpaa_ratingUnrated 8.311e-02 3.071e-01 0.271 0.78673
## imdb_num_votes -9.223e-07 4.300e-07 -2.145 0.03234 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.9816 on 620 degrees of freedom
## (1 observation deleted due to missingness)
## Multiple R-squared: 0.6522, Adjusted R-squared: 0.6399
## F-statistic: 52.85 on 22 and 620 DF, p-value: < 2.2e-16Remove highest p-value and run model again
MTM2<- lm(revenue_log ~ budget_log+facebook_likes+critics_score+audience_score+imdb_rating+runtime+mpaa_rating+imdb_num_votes, data = MT)
summary(MTM2)##
## Call:
## lm(formula = revenue_log ~ budget_log + facebook_likes + critics_score +
## audience_score + imdb_rating + runtime + mpaa_rating + imdb_num_votes,
## data = MT)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.0938 -0.5563 0.0415 0.6230 3.1233
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 2.477e-01 3.200e-01 0.774 0.43918
## budget_log 4.462e-01 3.488e-02 12.792 < 2e-16 ***
## facebook_likes 5.052e-06 4.582e-07 11.025 < 2e-16 ***
## critics_score 7.476e-03 2.684e-03 2.785 0.00552 **
## audience_score 3.074e-03 4.139e-03 0.743 0.45800
## imdb_rating 2.904e-02 4.345e-03 6.684 5.14e-11 ***
## runtime -3.305e-03 2.089e-03 -1.582 0.11416
## mpaa_ratingNC-17 6.642e-01 7.322e-01 0.907 0.36470
## mpaa_ratingPG 1.464e-01 2.453e-01 0.597 0.55083
## mpaa_ratingPG-13 3.071e-01 2.462e-01 1.248 0.21267
## mpaa_ratingR 1.471e-01 2.349e-01 0.626 0.53157
## mpaa_ratingUnrated -1.111e-01 2.690e-01 -0.413 0.67977
## imdb_num_votes -7.603e-07 4.236e-07 -1.795 0.07316 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.9813 on 630 degrees of freedom
## (1 observation deleted due to missingness)
## Multiple R-squared: 0.6468, Adjusted R-squared: 0.64
## F-statistic: 96.13 on 12 and 630 DF, p-value: < 2.2e-16MTM3<- lm(revenue_log ~ budget_log+facebook_likes+critics_score+audience_score+imdb_rating+runtime+imdb_num_votes, data = MT)
summary(MTM3)##
## Call:
## lm(formula = revenue_log ~ budget_log + facebook_likes + critics_score +
## audience_score + imdb_rating + runtime + imdb_num_votes,
## data = MT)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.3932 -0.5484 0.0351 0.6552 3.1412
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 3.620e-01 2.536e-01 1.427 0.1540
## budget_log 4.477e-01 3.490e-02 12.829 < 2e-16 ***
## facebook_likes 5.070e-06 4.572e-07 11.089 < 2e-16 ***
## critics_score 6.800e-03 2.664e-03 2.552 0.0109 *
## audience_score 2.032e-03 4.125e-03 0.493 0.6224
## imdb_rating 3.002e-02 4.333e-03 6.929 1.05e-11 ***
## runtime -2.617e-03 2.051e-03 -1.276 0.2024
## imdb_num_votes -6.548e-07 4.203e-07 -1.558 0.1198
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.9831 on 635 degrees of freedom
## (1 observation deleted due to missingness)
## Multiple R-squared: 0.6427, Adjusted R-squared: 0.6388
## F-statistic: 163.2 on 7 and 635 DF, p-value: < 2.2e-16MTM4<- lm(revenue_log ~ budget_log+facebook_likes+critics_score+imdb_rating+runtime+imdb_num_votes, data = MT)
summary(MTM4)##
## Call:
## lm(formula = revenue_log ~ budget_log + facebook_likes + critics_score +
## imdb_rating + runtime + imdb_num_votes, data = MT)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.3821 -0.5511 0.0382 0.6527 3.1495
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 3.584e-01 2.534e-01 1.414 0.15774
## budget_log 4.481e-01 3.487e-02 12.852 < 2e-16 ***
## facebook_likes 5.094e-06 4.545e-07 11.208 < 2e-16 ***
## critics_score 7.458e-03 2.304e-03 3.238 0.00127 **
## imdb_rating 3.142e-02 3.277e-03 9.588 < 2e-16 ***
## runtime -2.625e-03 2.050e-03 -1.281 0.20079
## imdb_num_votes -6.303e-07 4.171e-07 -1.511 0.13128
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.9825 on 636 degrees of freedom
## (1 observation deleted due to missingness)
## Multiple R-squared: 0.6426, Adjusted R-squared: 0.6392
## F-statistic: 190.6 on 6 and 636 DF, p-value: < 2.2e-16MTM5<- lm(revenue_log ~ budget_log+facebook_likes+critics_score+imdb_rating+imdb_num_votes, data = MT)
summary(MTM5)##
## Call:
## lm(formula = revenue_log ~ budget_log + facebook_likes + critics_score +
## imdb_rating + imdb_num_votes, data = MT)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.4224 -0.5445 0.0182 0.6441 3.1447
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 9.122e-02 1.420e-01 0.642 0.52087
## budget_log 4.492e-01 3.485e-02 12.889 < 2e-16 ***
## facebook_likes 5.087e-06 4.544e-07 11.196 < 2e-16 ***
## critics_score 7.316e-03 2.299e-03 3.182 0.00153 **
## imdb_rating 3.143e-02 3.274e-03 9.601 < 2e-16 ***
## imdb_num_votes -7.521e-07 4.063e-07 -1.851 0.06465 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.9823 on 638 degrees of freedom
## Multiple R-squared: 0.6416, Adjusted R-squared: 0.6388
## F-statistic: 228.4 on 5 and 638 DF, p-value: < 2.2e-16MTM6<- lm(revenue_log ~ budget_log+facebook_likes+critics_score+imdb_rating, data = MT)
summary(MTM6)##
## Call:
## lm(formula = revenue_log ~ budget_log + facebook_likes + critics_score +
## imdb_rating, data = MT)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.4278 -0.5570 0.0164 0.6507 3.2044
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.117e-01 1.419e-01 0.787 0.43132
## budget_log 4.511e-01 3.490e-02 12.923 < 2e-16 ***
## facebook_likes 5.090e-06 4.552e-07 11.183 < 2e-16 ***
## critics_score 7.211e-03 2.303e-03 3.132 0.00182 **
## imdb_rating 3.041e-02 3.233e-03 9.406 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.9842 on 639 degrees of freedom
## Multiple R-squared: 0.6397, Adjusted R-squared: 0.6374
## F-statistic: 283.6 on 4 and 639 DF, p-value: < 2.2e-16With final model we have all the variables that reject the null hypothesis
MTModelFinal <- lm(revenue_log ~ budget_log+facebook_likes+critics_score+imdb_rating, data = MT)
summary(MTModelFinal)##
## Call:
## lm(formula = revenue_log ~ budget_log + facebook_likes + critics_score +
## imdb_rating, data = MT)
##
## Residuals:
## Min 1Q Median 3Q Max
## -5.4278 -0.5570 0.0164 0.6507 3.2044
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.117e-01 1.419e-01 0.787 0.43132
## budget_log 4.511e-01 3.490e-02 12.923 < 2e-16 ***
## facebook_likes 5.090e-06 4.552e-07 11.183 < 2e-16 ***
## critics_score 7.211e-03 2.303e-03 3.132 0.00182 **
## imdb_rating 3.041e-02 3.233e-03 9.406 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.9842 on 639 degrees of freedom
## Multiple R-squared: 0.6397, Adjusted R-squared: 0.6374
## F-statistic: 283.6 on 4 and 639 DF, p-value: < 2.2e-16From summary Greatest Factors contributing to Revenue is Budget, Facebook likes, critic score and IMDB rating
Multiple R-squared: 0.64 indicates that 64% of the variation is explained by audience, critic, budget etc.
Reject null hypothesis for Budget, Facebook likes, critic score and IMDB rating as they show a relationship with revenue
Checking Correlation between variables to prevent Homoscedasticity
check_model(MTModelFinal)
View in table format for collinearity
performance::check_collinearity(MTModelFinal)## # Check for Multicollinearity
##
## Low Correlation
##
## Term VIF VIF 95% CI Increased SE Tolerance Tolerance 95% CI
## budget_log 1.34 [1.23, 1.49] 1.16 0.75 [0.67, 0.81]
## facebook_likes 1.21 [1.12, 1.35] 1.10 0.83 [0.74, 0.89]
## critics_score 1.88 [1.69, 2.12] 1.37 0.53 [0.47, 0.59]
## imdb_rating 2.10 [1.88, 2.38] 1.45 0.48 [0.42, 0.53]performance_accuracy(MTModelFinal)## # Accuracy of Model Predictions
##
## Accuracy (95% CI): 79.48% [76.04%, 83.84%]
## Method: Correlation between observed and predictedLook at Correlation
head(MT %>%
select(budget_log,revenue_log,critics_score, runtime, facebook_likes, imdb_rating),10)## budget_log revenue_log critics_score runtime facebook_likes imdb_rating
## 1 2.533697 2.9263822 60 134 23343 68
## 2 3.806662 4.0980045 5 108 84182 49
## 3 2.772589 3.4518905 40 97 35296 63
## 4 1.252763 -0.2876821 44 98 445 63
## 5 2.772589 2.9796029 51 111 21583 60
## 6 3.912023 5.4821378 75 106 20965 78
## 7 3.912023 3.5931942 35 87 12952 54
## 8 3.258097 3.7686144 50 83 52827 76
## 9 4.174387 5.4624325 70 100 48878 70
## 10 2.140066 1.7526721 53 86 5481 41MT_sub<-MT%>%select(budget_log,revenue_log,critics_score, facebook_likes, imdb_rating)
cor(MT_sub) # create the correlation matrix ## budget_log revenue_log critics_score facebook_likes imdb_rating
## budget_log 1.0000000 0.6319071 0.2844238 0.3738611 0.4292595
## revenue_log 0.6319071 1.0000000 0.5005811 0.5474557 0.6320976
## critics_score 0.2844238 0.5005811 1.0000000 0.2656832 0.6798900
## facebook_likes 0.3738611 0.5474557 0.2656832 1.0000000 0.2966853
## imdb_rating 0.4292595 0.6320976 0.6798900 0.2966853 1.0000000#anything over o.5 is highly correlated
# if - then negatively correlatedCompany screen a movie premiere that had a critic score of 55 and 1,250 likes on facebook
What is predicted box office sales?
head(predict(MTModelFinal),10)## 1 2 3 4 5 6 7 8
## 3.873931 3.783411 3.746259 2.912163 3.664545 4.895802 3.836724 4.521936
## 9 10
## 4.876898 2.733903Predict Revenue based on critic score and facebook likes
# Define new data for prediction (Critics score = 55, Facebook likes = 1,250)
Question_a <- data.frame(
budget_log = mean(MT$budget_log, na.rm = TRUE), # Default value
facebook_likes = 1250,
critics_score = 55,
audience_score = mean(MT$audience_score, na.rm = TRUE), # Default value
imdb_rating = mean(MT$imdb_rating, na.rm = TRUE), # Default value
runtime = mean(MT$runtime, na.rm = TRUE), # Default value
genre = levels(MT$genre)[1] # Default category
)Print Predicted Revenue with the log transformation not applied
pre_rev_a <- predict(MTModelFinal, newdata = Question_a)
# Residual variance from the log-transformed model
sigma_squared <- summary(MTModelFinal)$sigma^2
# Adjusted prediction back to the original scale
adjusted_prediction_a <- exp(pre_rev_a + sigma_squared / 2)
print(adjusted_prediction_a)## 1
## 61.23421Part B. if the producers conduct an advertsing campaign leadings to an overall movie budget of $20 million and is expected to increase Facebook likes be 10,000%. how is that campaign expected to improve overall revenue?
# Define new data for the campaign
Question_b <- data.frame(
budget_log = log(20),
facebook_likes = 126250, # Updated likes
critics_score = 55, # Same critics score
audience_score = mean(MT$audience_score, na.rm = TRUE), # Default value
imdb_rating = mean(MT$imdb_rating, na.rm = TRUE), # Default value
runtime = mean(MT$runtime, na.rm = TRUE), # Default value
genre = levels(MT$genre)[1] # Default category
)Print prediction without log transformation
pre_rev_b <- predict(MTModelFinal, newdata = Question_b)
# Residual variance from the log-transformed model
sigma_squared <- summary(MTModelFinal)$sigma^2
# Adjusted prediction back to the original scale
adjusted_prediction_b<- exp(pre_rev_b + sigma_squared / 2)
print(adjusted_prediction_b)## 1
## 111.7836Based on the predictive model we can expect an increase of
comp = adjusted_prediction_b - adjusted_prediction_a
print(comp)## 1
## 50.54937# or percentage increase of
percent_increase = ((adjusted_prediction_b - adjusted_prediction_a)/ adjusted_prediction_a) * 100
print(percent_increase)## 1
## 82.55087