Peer Graded Assignment: Prediction Assignment Writeup
Tadashi Horie
2016年6月28日
Overview
The goal of this project is to predict the manner in which they did the exercise. This is the “classe” variable in the training set. This report describes how data was cleaned, how I split “pml-training.csv” into train set and test set, and some of models are investigated.
Exercise
1. Loading add-on package and set seed
set.seed(12345)
library(caret)2. Download rawdata and submit_data
url_train <- "https://d396qusza40orc.cloudfront.net/predmachlearn/pml-training.csv"
rawdata <- read.csv(url_train, na.strings = c("", "NA"))
url_submit <- "https://d396qusza40orc.cloudfront.net/predmachlearn/pml-testing.csv"
submit_data <- read.csv(url_submit, na.strings = c("", "NA"))3. Cleaning data
We should delete the column that contains NA to avoid the error. In addition, in order to make accurate predictions, columns that is not related exercise must also be deleted. In particular “X”, “user_name”, “raw_timestamp_part_1”, “raw_timestamp_part_2”, “cvtd_timestamp”, “new_window”, “num_window” are deleted.
#Remove NA cols
colname <- colnames(rawdata)[!colSums(is.na(rawdata)) > 0]
colname## [1] "X" "user_name" "raw_timestamp_part_1"
## [4] "raw_timestamp_part_2" "cvtd_timestamp" "new_window"
## [7] "num_window" "roll_belt" "pitch_belt"
## [10] "yaw_belt" "total_accel_belt" "gyros_belt_x"
## [13] "gyros_belt_y" "gyros_belt_z" "accel_belt_x"
## [16] "accel_belt_y" "accel_belt_z" "magnet_belt_x"
## [19] "magnet_belt_y" "magnet_belt_z" "roll_arm"
## [22] "pitch_arm" "yaw_arm" "total_accel_arm"
## [25] "gyros_arm_x" "gyros_arm_y" "gyros_arm_z"
## [28] "accel_arm_x" "accel_arm_y" "accel_arm_z"
## [31] "magnet_arm_x" "magnet_arm_y" "magnet_arm_z"
## [34] "roll_dumbbell" "pitch_dumbbell" "yaw_dumbbell"
## [37] "total_accel_dumbbell" "gyros_dumbbell_x" "gyros_dumbbell_y"
## [40] "gyros_dumbbell_z" "accel_dumbbell_x" "accel_dumbbell_y"
## [43] "accel_dumbbell_z" "magnet_dumbbell_x" "magnet_dumbbell_y"
## [46] "magnet_dumbbell_z" "roll_forearm" "pitch_forearm"
## [49] "yaw_forearm" "total_accel_forearm" "gyros_forearm_x"
## [52] "gyros_forearm_y" "gyros_forearm_z" "accel_forearm_x"
## [55] "accel_forearm_y" "accel_forearm_z" "magnet_forearm_x"
## [58] "magnet_forearm_y" "magnet_forearm_z" "classe"#Slice data related with exercise
colname <- colname[8: length(colname)]
df_wo_NA <- rawdata[colname]
#Check the colnames of df_wo_NA is in submit_data.
#The last colname is "classe"
is.element(colname, colnames(submit_data))## [1] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [12] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [23] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [34] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE
## [45] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE FALSE4. Split data into random train and test
inTrain = createDataPartition(df_wo_NA$classe, p = 3/4)[[1]]
training = df_wo_NA[ inTrain,]
testing = df_wo_NA[-inTrain,]4. Random Forest
It takes a very long time for training, but it has a high accuracy.
model_rf <- train(classe ~ ., data = training, method = "rf")
pred_rf <- predict(model_rf, testing)
confusionMatrix(testing$classe, pred_rf)## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 1395 0 0 0 0
## B 9 934 6 0 0
## C 0 3 848 4 0
## D 0 0 3 801 0
## E 0 0 1 1 899
##
## Overall Statistics
##
## Accuracy : 0.9945
## 95% CI : (0.992, 0.9964)
## No Information Rate : 0.2863
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.993
## Mcnemar's Test P-Value : NA
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 0.9936 0.9968 0.9883 0.9938 1.0000
## Specificity 1.0000 0.9962 0.9983 0.9993 0.9995
## Pos Pred Value 1.0000 0.9842 0.9918 0.9963 0.9978
## Neg Pred Value 0.9974 0.9992 0.9975 0.9988 1.0000
## Prevalence 0.2863 0.1911 0.1750 0.1644 0.1833
## Detection Rate 0.2845 0.1905 0.1729 0.1633 0.1833
## Detection Prevalence 0.2845 0.1935 0.1743 0.1639 0.1837
## Balanced Accuracy 0.9968 0.9965 0.9933 0.9965 0.99985. Liner Discriminant Analysis
It takes a short time but poor accuracy.
model_lda <- train(classe ~ ., data = training, method = "lda")
pred_lda <- predict(model_lda, testing)
confusionMatrix(testing$classe, pred_lda)## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 1150 29 103 103 10
## B 159 589 125 34 42
## C 100 71 558 101 25
## D 34 40 89 609 32
## E 31 151 84 83 552
##
## Overall Statistics
##
## Accuracy : 0.7051
## 95% CI : (0.6922, 0.7179)
## No Information Rate : 0.3006
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.6267
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 0.7802 0.6693 0.5819 0.6548 0.8351
## Specificity 0.9286 0.9105 0.9247 0.9509 0.9177
## Pos Pred Value 0.8244 0.6207 0.6526 0.7575 0.6127
## Neg Pred Value 0.9077 0.9264 0.9010 0.9217 0.9728
## Prevalence 0.3006 0.1794 0.1956 0.1896 0.1348
## Detection Rate 0.2345 0.1201 0.1138 0.1242 0.1126
## Detection Prevalence 0.2845 0.1935 0.1743 0.1639 0.1837
## Balanced Accuracy 0.8544 0.7899 0.7533 0.8029 0.87646. Recursive Partitioning and Regression Trees
The results can be confirmed visually, but poor accuracy.
model_rpart <- train(classe ~ ., data = training, method = "rpart")
pred_rpart<- predict(model_rpart, testing)
confusionMatrix(testing$classe, pred_rpart)## Confusion Matrix and Statistics
##
## Reference
## Prediction A B C D E
## A 859 215 203 109 9
## B 151 434 166 197 1
## C 27 117 442 269 0
## D 43 65 131 492 73
## E 14 143 111 133 500
##
## Overall Statistics
##
## Accuracy : 0.5561
## 95% CI : (0.542, 0.57)
## No Information Rate : 0.2447
## P-Value [Acc > NIR] : < 2.2e-16
##
## Kappa : 0.4442
## Mcnemar's Test P-Value : < 2.2e-16
##
## Statistics by Class:
##
## Class: A Class: B Class: C Class: D Class: E
## Sensitivity 0.7852 0.4456 0.41975 0.4100 0.8576
## Specificity 0.8593 0.8690 0.89276 0.9158 0.9072
## Pos Pred Value 0.6158 0.4573 0.51696 0.6119 0.5549
## Neg Pred Value 0.9330 0.8635 0.84910 0.8273 0.9793
## Prevalence 0.2231 0.1986 0.21472 0.2447 0.1189
## Detection Rate 0.1752 0.0885 0.09013 0.1003 0.1020
## Detection Prevalence 0.2845 0.1935 0.17435 0.1639 0.1837
## Balanced Accuracy 0.8223 0.6573 0.65625 0.6629 0.8824library(rattle)
fancyRpartPlot(model_rpart$finalModel)
7. Submit data with Random Forest
We can use the high accuracy model to submit data. In this report the Random Forest accuracy has the highest value 99.45. We can show the prediction.
submit_rf <- predict(model_rf, submit_data)
submit_rf## [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 E