## Executive Summary
Using accelerometer data from the belt, forearm, arm, and dumbbell of 6 participants performing barbell lifts correctly and incorrectly in 5 ways (classes A-E), we built a random forest model to predict the manner in which an exercise was performed. After cleaning the data and holding out 30% for validation, the model achieved about 99% validation accuracy.
library(caret)
library(randomForest)
training_raw <- read.csv("pml-training.csv", na.strings = c("NA", "", "#DIV/0!"))
testing_raw <- read.csv("pml-testing.csv", na.strings = c("NA", "", "#DIV/0!"))
na_prop <- colMeans(is.na(training_raw))
training_clean <- training_raw[, na_prop < 0.9]
nzv <- nearZeroVar(training_clean)
training_clean <- training_clean[, -nzv]
id_cols <- grep("^X$|user_name|timestamp|window", names(training_clean))
training_clean <- training_clean[, -id_cols]
training_clean$classe <- factor(training_clean$classe)
dim(training_clean)## [1] 19622 53set.seed(12345)
inTrain <- createDataPartition(training_clean$classe, p = 0.7, list = FALSE)
trainSet <- training_clean[inTrain, ]
validSet <- training_clean[-inTrain, ]ctrl <- trainControl(method = "cv", number = 5)
set.seed(12345)
modFit <- train(classe ~ ., data = trainSet, method = "rf",
trControl = ctrl, ntree = 150, importance = TRUE)
modFit## Random Forest
##
## 13737 samples
## 52 predictor
## 5 classes: 'A', 'B', 'C', 'D', 'E'
##
## No pre-processing
## Resampling: Cross-Validated (5 fold)
## Summary of sample sizes: 10990, 10990, 10989, 10991, 10988
## Resampling results across tuning parameters:
##
## mtry Accuracy Kappa
## 2 0.9916284 0.9894090
## 27 0.9908273 0.9883964
## 52 0.9839117 0.9796484
##
## Accuracy was used to select the optimal model using the largest value.
## The final value used for the model was mtry = 2.predValid <- predict(modFit, validSet)
confusionMatrix(predValid, validSet$classe)## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 1673 9 0 0 0
## B 1 1130 6 0 0
## C 0 0 1020 25 1
## D 0 0 0 937 0
## E 0 0 0 2 1081
##
## Overall Statistics
##
## Accuracy : 0.9925
## 95% CI : (0.99, 0.9946)
## No Information Rate : 0.2845
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9905
##
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 0.9994 0.9921 0.9942 0.9720 0.9991
## Specificity 0.9979 0.9985 0.9946 1.0000 0.9996
## Pos Pred Value 0.9946 0.9938 0.9751 1.0000 0.9982
## Neg Pred Value 0.9998 0.9981 0.9988 0.9945 0.9998
## Prevalence 0.2845 0.1935 0.1743 0.1638 0.1839
## Detection Rate 0.2843 0.1920 0.1733 0.1592 0.1837
## Detection Prevalence 0.2858 0.1932 0.1777 0.1592 0.1840
## Balanced Accuracy 0.9986 0.9953 0.9944 0.9860 0.9993predTest <- predict(modFit, testing_raw)
data.frame(problem_id = testing_raw$problem_id, prediction = predTest)## problem_id prediction
## 1 1 B
## 2 2 A
## 3 3 B
## 4 4 A
## 5 5 A
## 6 6 E
## 7 7 D
## 8 8 B
## 9 9 A
## 10 10 A
## 11 11 B
## 12 12 C
## 13 13 B
## 14 14 A
## 15 15 E
## 16 16 E
## 17 17 A
## 18 18 B
## 19 19 B
## 20 20 BThe random forest model achieved approximately 99% accuracy on the held-out validation set, giving high confidence in the final predictions above.