This is a group project and we use Oscar Academy Awards data to determine whether Best Film Editing is the best predictor of Best Picture.
The input data is from link. The 1’s indicate winning the award, 0’s indicate not winning.
library(XML)
oscar_data <- readHTMLTable('tidedoscarwinners.csv.html')
oscar_data <- oscar_data[[1]]
head(oscar_data)## structure(c("1", "2", "3", "4", "5", "6"), class = "AsIs") Year
## 1 1 1934
## 2 2 1934
## 3 3 1934
## 4 4 1934
## 5 5 1935
## 6 6 1935
## pictures Best_Picture Best_Editing Best_Directing
## 1 Eskimo 0 1 0
## 2 It Happened One Night 1 0 1
## 3 One Night of Love 0 0 0
## 4 The Gay Divorcee 0 0 0
## 5 A Midsummer Night's Dream 0 1 0
## 6 Dangerous 0 0 0
## Best_Actor Best_Supporting_Actor Best_Actress Best_Supporting_Actress
## 1 0 0 0 0
## 2 1 0 1 0
## 3 0 0 0 0
## 4 0 0 0 0
## 5 0 0 0 0
## 6 0 0 1 0
## Best_Sound Best_Song
## 1 0 0
## 2 0 0
## 3 1 0
## 4 0 1
## 5 0 0
## 6 0 0oscar_data_ml <- oscar_data[,-c(1:4)]
head(oscar_data_ml)## Best_Picture Best_Editing Best_Directing Best_Actor
## 1 0 1 0 0
## 2 1 0 1 1
## 3 0 0 0 0
## 4 0 0 0 0
## 5 0 1 0 0
## 6 0 0 0 0
## Best_Supporting_Actor Best_Actress Best_Supporting_Actress Best_Sound
## 1 0 0 0 0
## 2 0 1 0 0
## 3 0 0 0 1
## 4 0 0 0 0
## 5 0 0 0 0
## 6 0 1 0 0
## Best_Song
## 1 0
## 2 0
## 3 0
## 4 1
## 5 0
## 6 0dim(oscar_data_ml)## [1] 457 9str(oscar_data_ml)## 'data.frame': 457 obs. of 9 variables:
## $ Best_Picture : Factor w/ 2 levels "0","1": 1 2 1 1 1 1 1 2 1 1 ...
## $ Best_Editing : Factor w/ 2 levels "0","1": 2 1 1 1 2 1 1 1 1 1 ...
## $ Best_Directing : Factor w/ 2 levels "0","1": 1 2 1 1 1 1 1 1 1 2 ...
## $ Best_Actor : Factor w/ 2 levels "0","1": 1 2 1 1 1 1 1 1 1 2 ...
## $ Best_Supporting_Actor : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
## $ Best_Actress : Factor w/ 2 levels "0","1": 1 2 1 1 1 2 1 1 1 1 ...
## $ Best_Supporting_Actress: Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
## $ Best_Sound : Factor w/ 2 levels "0","1": 1 1 2 1 1 1 1 1 2 1 ...
## $ Best_Song : Factor w/ 2 levels "0","1": 1 1 1 2 1 1 2 1 1 1 ...library(lattice)
library(ggplot2)
library(caret)
set.seed(45)
inTrain <- createDataPartition(y=oscar_data_ml$Best_Picture, p=0.7,list=FALSE)
training <- oscar_data_ml[inTrain,]
testing <- oscar_data_ml[-inTrain,]
dim(training)## [1] 321 9dim(testing)## [1] 136 9library(randomForest)## randomForest 4.6-10
## Type rfNews() to see new features/changes/bug fixes.model <- randomForest(Best_Picture ~., data=training)
model##
## Call:
## randomForest(formula = Best_Picture ~ ., data = training)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 2
##
## OOB estimate of error rate: 8.41%
## Confusion matrix:
## 0 1 class.error
## 0 251 13 0.04924242
## 1 14 43 0.24561404importance(model)## MeanDecreaseGini
## Best_Editing 4.2596672
## Best_Directing 33.5250633
## Best_Actor 2.7116693
## Best_Supporting_Actor 1.0599584
## Best_Actress 1.0723682
## Best_Supporting_Actress 0.9064921
## Best_Sound 1.9436359
## Best_Song 1.8514441library(caret)
predicted <- predict(model, testing)
table(predicted)## predicted
## 0 1
## 112 24library(e1071)
confusionMatrix(predicted, testing$Best_Picture)## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 106 6
## 1 6 18
##
## Accuracy : 0.9118
## 95% CI : (0.8509, 0.9536)
## No Information Rate : 0.8235
## P-Value [Acc > NIR] : 0.002799
##
## Kappa : 0.6964
## Mcnemar's Test P-Value : 1.000000
##
## Sensitivity : 0.9464
## Specificity : 0.7500
## Pos Pred Value : 0.9464
## Neg Pred Value : 0.7500
## Prevalence : 0.8235
## Detection Rate : 0.7794
## Detection Prevalence : 0.8235
## Balanced Accuracy : 0.8482
##
## 'Positive' Class : 0
## With this dataset, Random forests provide a nice model to predict whether movie wins Best Picture or not. From the model, we can see Best_Directing is the strongest predictor, and Best_Editing is the next.