PML Final Project
Jose Pereira
17 de Outubro de 2015
Background
Using devices such as Jawbone Up, Nike FuelBand, and Fitbit it is now possible to collect a large amount of data about personal activity relatively inexpensively. These type of devices are part of the quantified self movement – a group of enthusiasts who take measurements about themselves regularly to improve their health, to find patterns in their behavior, or because they are tech geeks. One thing that people regularly do is quantify how much of a particular activity they do, but they rarely quantify how well they do it. In this project, your goal will be to use data from accelerometers on the belt, forearm, arm, and dumbell of 6 participants. They were asked to perform barbell lifts correctly and incorrectly in 5 different ways. More information is available from the website here: http://groupware.les.inf.puc-rio.br/har (see the section on the Weight Lifting Exercise Dataset).
Goal of the project
The goal of your project is to predict the manner in which they did the exercise. This is the “classe” variable in the training set. You may use any of the other variables to predict with. You should create a report describing how you built your model, how you used cross validation, what you think the expected out of sample error is, and why you made the choices you did. You will also use your prediction model to predict 20 different test cases.
Data Loading and Cleaning
I will first start by loading the data and defining all blank and badly coded variables as N.A
train_data <- read.csv("pml-training.csv", sep=",", na.strings = c("", " ", "#DIV/0!","NA"))
test_data <- read.csv("pml-testing.csv", sep="," ,na.strings = c("", " ", "#DIV/0!","NA"))I will then check the structure of both datasets
str(train_data)## 'data.frame': 19622 obs. of 160 variables:
## $ X : int 1 2 3 4 5 6 7 8 9 10 ...
## $ user_name : Factor w/ 6 levels "adelmo","carlitos",..: 2 2 2 2 2 2 2 2 2 2 ...
## $ raw_timestamp_part_1 : int 1323084231 1323084231 1323084231 1323084232 1323084232 1323084232 1323084232 1323084232 1323084232 1323084232 ...
## $ raw_timestamp_part_2 : int 788290 808298 820366 120339 196328 304277 368296 440390 484323 484434 ...
## $ cvtd_timestamp : Factor w/ 20 levels "02/12/2011 13:32",..: 9 9 9 9 9 9 9 9 9 9 ...
## $ new_window : Factor w/ 2 levels "no","yes": 1 1 1 1 1 1 1 1 1 1 ...
## $ num_window : int 11 11 11 12 12 12 12 12 12 12 ...
## $ roll_belt : num 1.41 1.41 1.42 1.48 1.48 1.45 1.42 1.42 1.43 1.45 ...
## $ pitch_belt : num 8.07 8.07 8.07 8.05 8.07 8.06 8.09 8.13 8.16 8.17 ...
## $ yaw_belt : num -94.4 -94.4 -94.4 -94.4 -94.4 -94.4 -94.4 -94.4 -94.4 -94.4 ...
## $ total_accel_belt : int 3 3 3 3 3 3 3 3 3 3 ...
## $ kurtosis_roll_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ kurtosis_picth_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ kurtosis_yaw_belt : logi NA NA NA NA NA NA ...
## $ skewness_roll_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ skewness_roll_belt.1 : num NA NA NA NA NA NA NA NA NA NA ...
## $ skewness_yaw_belt : logi NA NA NA NA NA NA ...
## $ max_roll_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ max_picth_belt : int NA NA NA NA NA NA NA NA NA NA ...
## $ max_yaw_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ min_roll_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ min_pitch_belt : int NA NA NA NA NA NA NA NA NA NA ...
## $ min_yaw_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ amplitude_roll_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ amplitude_pitch_belt : int NA NA NA NA NA NA NA NA NA NA ...
## $ amplitude_yaw_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ var_total_accel_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ avg_roll_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ stddev_roll_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ var_roll_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ avg_pitch_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ stddev_pitch_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ var_pitch_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ avg_yaw_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ stddev_yaw_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ var_yaw_belt : num NA NA NA NA NA NA NA NA NA NA ...
## $ gyros_belt_x : num 0 0.02 0 0.02 0.02 0.02 0.02 0.02 0.02 0.03 ...
## $ gyros_belt_y : num 0 0 0 0 0.02 0 0 0 0 0 ...
## $ gyros_belt_z : num -0.02 -0.02 -0.02 -0.03 -0.02 -0.02 -0.02 -0.02 -0.02 0 ...
## $ accel_belt_x : int -21 -22 -20 -22 -21 -21 -22 -22 -20 -21 ...
## $ accel_belt_y : int 4 4 5 3 2 4 3 4 2 4 ...
## $ accel_belt_z : int 22 22 23 21 24 21 21 21 24 22 ...
## $ magnet_belt_x : int -3 -7 -2 -6 -6 0 -4 -2 1 -3 ...
## $ magnet_belt_y : int 599 608 600 604 600 603 599 603 602 609 ...
## $ magnet_belt_z : int -313 -311 -305 -310 -302 -312 -311 -313 -312 -308 ...
## $ roll_arm : num -128 -128 -128 -128 -128 -128 -128 -128 -128 -128 ...
## $ pitch_arm : num 22.5 22.5 22.5 22.1 22.1 22 21.9 21.8 21.7 21.6 ...
## $ yaw_arm : num -161 -161 -161 -161 -161 -161 -161 -161 -161 -161 ...
## $ total_accel_arm : int 34 34 34 34 34 34 34 34 34 34 ...
## $ var_accel_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ avg_roll_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ stddev_roll_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ var_roll_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ avg_pitch_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ stddev_pitch_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ var_pitch_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ avg_yaw_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ stddev_yaw_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ var_yaw_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ gyros_arm_x : num 0 0.02 0.02 0.02 0 0.02 0 0.02 0.02 0.02 ...
## $ gyros_arm_y : num 0 -0.02 -0.02 -0.03 -0.03 -0.03 -0.03 -0.02 -0.03 -0.03 ...
## $ gyros_arm_z : num -0.02 -0.02 -0.02 0.02 0 0 0 0 -0.02 -0.02 ...
## $ accel_arm_x : int -288 -290 -289 -289 -289 -289 -289 -289 -288 -288 ...
## $ accel_arm_y : int 109 110 110 111 111 111 111 111 109 110 ...
## $ accel_arm_z : int -123 -125 -126 -123 -123 -122 -125 -124 -122 -124 ...
## $ magnet_arm_x : int -368 -369 -368 -372 -374 -369 -373 -372 -369 -376 ...
## $ magnet_arm_y : int 337 337 344 344 337 342 336 338 341 334 ...
## $ magnet_arm_z : int 516 513 513 512 506 513 509 510 518 516 ...
## $ kurtosis_roll_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ kurtosis_picth_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ kurtosis_yaw_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ skewness_roll_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ skewness_pitch_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ skewness_yaw_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ max_roll_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ max_picth_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ max_yaw_arm : int NA NA NA NA NA NA NA NA NA NA ...
## $ min_roll_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ min_pitch_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ min_yaw_arm : int NA NA NA NA NA NA NA NA NA NA ...
## $ amplitude_roll_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ amplitude_pitch_arm : num NA NA NA NA NA NA NA NA NA NA ...
## $ amplitude_yaw_arm : int NA NA NA NA NA NA NA NA NA NA ...
## $ roll_dumbbell : num 13.1 13.1 12.9 13.4 13.4 ...
## $ pitch_dumbbell : num -70.5 -70.6 -70.3 -70.4 -70.4 ...
## $ yaw_dumbbell : num -84.9 -84.7 -85.1 -84.9 -84.9 ...
## $ kurtosis_roll_dumbbell : num NA NA NA NA NA NA NA NA NA NA ...
## $ kurtosis_picth_dumbbell : num NA NA NA NA NA NA NA NA NA NA ...
## $ kurtosis_yaw_dumbbell : logi NA NA NA NA NA NA ...
## $ skewness_roll_dumbbell : num NA NA NA NA NA NA NA NA NA NA ...
## $ skewness_pitch_dumbbell : num NA NA NA NA NA NA NA NA NA NA ...
## $ skewness_yaw_dumbbell : logi NA NA NA NA NA NA ...
## $ max_roll_dumbbell : num NA NA NA NA NA NA NA NA NA NA ...
## $ max_picth_dumbbell : num NA NA NA NA NA NA NA NA NA NA ...
## $ max_yaw_dumbbell : num NA NA NA NA NA NA NA NA NA NA ...
## $ min_roll_dumbbell : num NA NA NA NA NA NA NA NA NA NA ...
## $ min_pitch_dumbbell : num NA NA NA NA NA NA NA NA NA NA ...
## $ min_yaw_dumbbell : num NA NA NA NA NA NA NA NA NA NA ...
## $ amplitude_roll_dumbbell : num NA NA NA NA NA NA NA NA NA NA ...
## [list output truncated]str(test_data)## 'data.frame': 20 obs. of 160 variables:
## $ X : int 1 2 3 4 5 6 7 8 9 10 ...
## $ user_name : Factor w/ 6 levels "adelmo","carlitos",..: 6 5 5 1 4 5 5 5 2 3 ...
## $ raw_timestamp_part_1 : int 1323095002 1322673067 1322673075 1322832789 1322489635 1322673149 1322673128 1322673076 1323084240 1322837822 ...
## $ raw_timestamp_part_2 : int 868349 778725 342967 560311 814776 510661 766645 54671 916313 384285 ...
## $ cvtd_timestamp : Factor w/ 11 levels "02/12/2011 13:33",..: 5 10 10 1 6 11 11 10 3 2 ...
## $ new_window : Factor w/ 1 level "no": 1 1 1 1 1 1 1 1 1 1 ...
## $ num_window : int 74 431 439 194 235 504 485 440 323 664 ...
## $ roll_belt : num 123 1.02 0.87 125 1.35 -5.92 1.2 0.43 0.93 114 ...
## $ pitch_belt : num 27 4.87 1.82 -41.6 3.33 1.59 4.44 4.15 6.72 22.4 ...
## $ yaw_belt : num -4.75 -88.9 -88.5 162 -88.6 -87.7 -87.3 -88.5 -93.7 -13.1 ...
## $ total_accel_belt : int 20 4 5 17 3 4 4 4 4 18 ...
## $ kurtosis_roll_belt : logi NA NA NA NA NA NA ...
## $ kurtosis_picth_belt : logi NA NA NA NA NA NA ...
## $ kurtosis_yaw_belt : logi NA NA NA NA NA NA ...
## $ skewness_roll_belt : logi NA NA NA NA NA NA ...
## $ skewness_roll_belt.1 : logi NA NA NA NA NA NA ...
## $ skewness_yaw_belt : logi NA NA NA NA NA NA ...
## $ max_roll_belt : logi NA NA NA NA NA NA ...
## $ max_picth_belt : logi NA NA NA NA NA NA ...
## $ max_yaw_belt : logi NA NA NA NA NA NA ...
## $ min_roll_belt : logi NA NA NA NA NA NA ...
## $ min_pitch_belt : logi NA NA NA NA NA NA ...
## $ min_yaw_belt : logi NA NA NA NA NA NA ...
## $ amplitude_roll_belt : logi NA NA NA NA NA NA ...
## $ amplitude_pitch_belt : logi NA NA NA NA NA NA ...
## $ amplitude_yaw_belt : logi NA NA NA NA NA NA ...
## $ var_total_accel_belt : logi NA NA NA NA NA NA ...
## $ avg_roll_belt : logi NA NA NA NA NA NA ...
## $ stddev_roll_belt : logi NA NA NA NA NA NA ...
## $ var_roll_belt : logi NA NA NA NA NA NA ...
## $ avg_pitch_belt : logi NA NA NA NA NA NA ...
## $ stddev_pitch_belt : logi NA NA NA NA NA NA ...
## $ var_pitch_belt : logi NA NA NA NA NA NA ...
## $ avg_yaw_belt : logi NA NA NA NA NA NA ...
## $ stddev_yaw_belt : logi NA NA NA NA NA NA ...
## $ var_yaw_belt : logi NA NA NA NA NA NA ...
## $ gyros_belt_x : num -0.5 -0.06 0.05 0.11 0.03 0.1 -0.06 -0.18 0.1 0.14 ...
## $ gyros_belt_y : num -0.02 -0.02 0.02 0.11 0.02 0.05 0 -0.02 0 0.11 ...
## $ gyros_belt_z : num -0.46 -0.07 0.03 -0.16 0 -0.13 0 -0.03 -0.02 -0.16 ...
## $ accel_belt_x : int -38 -13 1 46 -8 -11 -14 -10 -15 -25 ...
## $ accel_belt_y : int 69 11 -1 45 4 -16 2 -2 1 63 ...
## $ accel_belt_z : int -179 39 49 -156 27 38 35 42 32 -158 ...
## $ magnet_belt_x : int -13 43 29 169 33 31 50 39 -6 10 ...
## $ magnet_belt_y : int 581 636 631 608 566 638 622 635 600 601 ...
## $ magnet_belt_z : int -382 -309 -312 -304 -418 -291 -315 -305 -302 -330 ...
## $ roll_arm : num 40.7 0 0 -109 76.1 0 0 0 -137 -82.4 ...
## $ pitch_arm : num -27.8 0 0 55 2.76 0 0 0 11.2 -63.8 ...
## $ yaw_arm : num 178 0 0 -142 102 0 0 0 -167 -75.3 ...
## $ total_accel_arm : int 10 38 44 25 29 14 15 22 34 32 ...
## $ var_accel_arm : logi NA NA NA NA NA NA ...
## $ avg_roll_arm : logi NA NA NA NA NA NA ...
## $ stddev_roll_arm : logi NA NA NA NA NA NA ...
## $ var_roll_arm : logi NA NA NA NA NA NA ...
## $ avg_pitch_arm : logi NA NA NA NA NA NA ...
## $ stddev_pitch_arm : logi NA NA NA NA NA NA ...
## $ var_pitch_arm : logi NA NA NA NA NA NA ...
## $ avg_yaw_arm : logi NA NA NA NA NA NA ...
## $ stddev_yaw_arm : logi NA NA NA NA NA NA ...
## $ var_yaw_arm : logi NA NA NA NA NA NA ...
## $ gyros_arm_x : num -1.65 -1.17 2.1 0.22 -1.96 0.02 2.36 -3.71 0.03 0.26 ...
## $ gyros_arm_y : num 0.48 0.85 -1.36 -0.51 0.79 0.05 -1.01 1.85 -0.02 -0.5 ...
## $ gyros_arm_z : num -0.18 -0.43 1.13 0.92 -0.54 -0.07 0.89 -0.69 -0.02 0.79 ...
## $ accel_arm_x : int 16 -290 -341 -238 -197 -26 99 -98 -287 -301 ...
## $ accel_arm_y : int 38 215 245 -57 200 130 79 175 111 -42 ...
## $ accel_arm_z : int 93 -90 -87 6 -30 -19 -67 -78 -122 -80 ...
## $ magnet_arm_x : int -326 -325 -264 -173 -170 396 702 535 -367 -420 ...
## $ magnet_arm_y : int 385 447 474 257 275 176 15 215 335 294 ...
## $ magnet_arm_z : int 481 434 413 633 617 516 217 385 520 493 ...
## $ kurtosis_roll_arm : logi NA NA NA NA NA NA ...
## $ kurtosis_picth_arm : logi NA NA NA NA NA NA ...
## $ kurtosis_yaw_arm : logi NA NA NA NA NA NA ...
## $ skewness_roll_arm : logi NA NA NA NA NA NA ...
## $ skewness_pitch_arm : logi NA NA NA NA NA NA ...
## $ skewness_yaw_arm : logi NA NA NA NA NA NA ...
## $ max_roll_arm : logi NA NA NA NA NA NA ...
## $ max_picth_arm : logi NA NA NA NA NA NA ...
## $ max_yaw_arm : logi NA NA NA NA NA NA ...
## $ min_roll_arm : logi NA NA NA NA NA NA ...
## $ min_pitch_arm : logi NA NA NA NA NA NA ...
## $ min_yaw_arm : logi NA NA NA NA NA NA ...
## $ amplitude_roll_arm : logi NA NA NA NA NA NA ...
## $ amplitude_pitch_arm : logi NA NA NA NA NA NA ...
## $ amplitude_yaw_arm : logi NA NA NA NA NA NA ...
## $ roll_dumbbell : num -17.7 54.5 57.1 43.1 -101.4 ...
## $ pitch_dumbbell : num 25 -53.7 -51.4 -30 -53.4 ...
## $ yaw_dumbbell : num 126.2 -75.5 -75.2 -103.3 -14.2 ...
## $ kurtosis_roll_dumbbell : logi NA NA NA NA NA NA ...
## $ kurtosis_picth_dumbbell : logi NA NA NA NA NA NA ...
## $ kurtosis_yaw_dumbbell : logi NA NA NA NA NA NA ...
## $ skewness_roll_dumbbell : logi NA NA NA NA NA NA ...
## $ skewness_pitch_dumbbell : logi NA NA NA NA NA NA ...
## $ skewness_yaw_dumbbell : logi NA NA NA NA NA NA ...
## $ max_roll_dumbbell : logi NA NA NA NA NA NA ...
## $ max_picth_dumbbell : logi NA NA NA NA NA NA ...
## $ max_yaw_dumbbell : logi NA NA NA NA NA NA ...
## $ min_roll_dumbbell : logi NA NA NA NA NA NA ...
## $ min_pitch_dumbbell : logi NA NA NA NA NA NA ...
## $ min_yaw_dumbbell : logi NA NA NA NA NA NA ...
## $ amplitude_roll_dumbbell : logi NA NA NA NA NA NA ...
## [list output truncated]Since some NAs exist in the data in quite an abundant quantity, I decided to remove those columns that have more than 80% NAs
train_data <- train_data[,colSums(is.na(train_data)) < nrow(train_data) * 0.8]
test_data <- test_data[,colSums(is.na(test_data)) < nrow(test_data) * 0.8]Let’s remove the columns 1:5 which have data that seems to have no value for making a prediction model.
train_data <- train_data[,-(1:5)]
test_data <- test_data[,-(1:5)]Model building
In this classification problem, we are mainly worried with accuracy and not the interpretability of the model so I decided to just implement a Random Forest algorithm which is one of the most accurate and efficient ones instead of doing classification trees or boosting algorithms. I will start by setting a seed and loading the caret package and splitting the train_data in a proportion of 70% (the training data) and 30% (a validation set to test the data before applying it to the test data given by professor Leek).
set.seed(19)
library(caret)## Warning: package 'caret' was built under R version 3.2.2## Loading required package: lattice
## Loading required package: ggplot2in_train <- createDataPartition(y=train_data$classe, p = 0.7, list=FALSE)
training <- train_data[in_train,]
inside_testing <- train_data[-in_train,]I will now build a Random Forest model with all variables as predictors using a 4-fold cross validation (will be quicker and Random Forest algorithm doesn’t need a really decent cross-validation as it is very resillient by itself)
set.seed(14)
nf <- trainControl(method="cv", number = 4)
RFmodel <- train(classe ~ .,
data=training,
method="rf",
importance = TRUE,
trControl = nf)## Loading required package: randomForest
## randomForest 4.6-10
## Type rfNews() to see new features/changes/bug fixes.Let us check the model data, its accuracy and error rate and make a plot of the model. We will also test its accuracy on the validation test data
RFmodel$finalModel##
## Call:
## randomForest(x = x, y = y, mtry = param$mtry, importance = TRUE)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 28
##
## OOB estimate of error rate: 0.21%
## Confusion matrix:
## A B C D E class.error
## A 3905 0 0 0 1 0.0002560164
## B 5 2650 3 0 0 0.0030097818
## C 0 3 2393 0 0 0.0012520868
## D 0 0 9 2242 1 0.0044404973
## E 0 0 0 7 2518 0.0027722772plot(RFmodel$finalModel)
predRF <- predict(RFmodel, inside_testing)
confusionMatrix(predRF, inside_testing$classe)## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 1674 1 0 0 0
## B 0 1138 6 0 0
## C 0 0 1020 3 0
## D 0 0 0 961 4
## E 0 0 0 0 1078
##
## Overall Statistics
##
## Accuracy : 0.9976
## 95% CI : (0.996, 0.9987)
## No Information Rate : 0.2845
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.997
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 1.0000 0.9991 0.9942 0.9969 0.9963
## Specificity 0.9998 0.9987 0.9994 0.9992 1.0000
## Pos Pred Value 0.9994 0.9948 0.9971 0.9959 1.0000
## Neg Pred Value 1.0000 0.9998 0.9988 0.9994 0.9992
## Prevalence 0.2845 0.1935 0.1743 0.1638 0.1839
## Detection Rate 0.2845 0.1934 0.1733 0.1633 0.1832
## Detection Prevalence 0.2846 0.1944 0.1738 0.1640 0.1832
## Balanced Accuracy 0.9999 0.9989 0.9968 0.9980 0.9982As we can see, the accuracy of this model on the training data validation set is quite good and its out of sample error rate is very low. Let us try to see the predictions in the test data of professor Leek.
final_predictions_perfect <- predict(RFmodel, test_data)## Loading required package: randomForest
## randomForest 4.6-10
## Type rfNews() to see new features/changes/bug fixes.final_predictions_perfect## [1] B A B A A E D B A A B C B A E E A B B B
## Levels: A B C D EConsidering the accuracy of this model, the only thing we can do now is to try to make the model creation faster by doing some feature editing and tuning some parameters. Let’s start by creating a plot of variable importance.
varImpPlot(RFmodel$finalModel, type=2)
We can see that some variables have more importance in the prediction than others. From the analysis of the past model, we know the perfect mtry is 27 and so we can try to tune nodesize from the default of 1 to a bigger value and try to diminish the number of trees a bit to get a faster model using only the most important variables.
tune <- expand.grid(mtry = 28)
RFmodel2 <- train(classe ~ num_window + roll_belt + pitch_forearm + yaw_belt + magnet_dumbbell_z + magnet_dumbbell_y + pitch_belt,
data = training,
method = "rf",
tuneGrid = tune,
ntree = 250, nodesize = 50)I will now test the data on both the validation set and the test data of professor Leek to compare the results with the slower model.
RFmodel2$finalModel##
## Call:
## randomForest(x = x, y = y, ntree = 250, mtry = param$mtry, nodesize = 50)
## Type of random forest: classification
## Number of trees: 250
## No. of variables tried at each split: 7
##
## OOB estimate of error rate: 2.1%
## Confusion matrix:
## A B C D E class.error
## A 3885 19 1 0 1 0.005376344
## B 19 2579 33 12 15 0.029721595
## C 3 33 2322 29 9 0.030884808
## D 4 1 21 2219 7 0.014653641
## E 6 34 22 20 2443 0.032475248RFmodel2## Random Forest
##
## 13737 samples
## 54 predictor
## 5 classes: 'A', 'B', 'C', 'D', 'E'
##
## No pre-processing
## Resampling: Bootstrapped (25 reps)
## Summary of sample sizes: 13737, 13737, 13737, 13737, 13737, 13737, ...
## Resampling results
##
## Accuracy Kappa Accuracy SD Kappa SD
## 0.9778785 0.9720128 0.004315469 0.005478715
##
## Tuning parameter 'mtry' was held constant at a value of 27
## predictions_inside_testing <- predict(RFmodel2, inside_testing)
confusionMatrix(predictions_inside_testing, inside_testing$classe)## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 1668 4 0 5 3
## B 6 1119 18 0 8
## C 0 11 990 9 4
## D 0 1 15 950 6
## E 0 4 3 0 1061
##
## Overall Statistics
##
## Accuracy : 0.9835
## 95% CI : (0.9799, 0.9866)
## No Information Rate : 0.2845
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.9791
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 0.9964 0.9824 0.9649 0.9855 0.9806
## Specificity 0.9972 0.9933 0.9951 0.9955 0.9985
## Pos Pred Value 0.9929 0.9722 0.9763 0.9774 0.9934
## Neg Pred Value 0.9986 0.9958 0.9926 0.9972 0.9956
## Prevalence 0.2845 0.1935 0.1743 0.1638 0.1839
## Detection Rate 0.2834 0.1901 0.1682 0.1614 0.1803
## Detection Prevalence 0.2855 0.1956 0.1723 0.1652 0.1815
## Balanced Accuracy 0.9968 0.9878 0.9800 0.9905 0.9896final_predictions_faster <- predict(RFmodel2, test_data)
final_predictions_perfect## [1] B A B A A E D B A A B C B A E E A B B B
## Levels: A B C D Efinal_predictions_faster## [1] B A B A A E D B A A B C B A E E A B B B
## Levels: A B C D EWe can see the accuracy and out of sample error of this model is not that different from the slower one and the test predictions are as accurate as those from the slower model, while being much faster to build. This proves how important it is to fine-tune the built models in order to gain the perfect balance between accuracy and speed.