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, union
batting <- read.csv("/Users/justincarter/Downloads/Batting.csv")
HOF <- read.csv("/Users/justincarter/Downloads/HallOfFame.csv")
fielding <- read.csv("/Users/justincarter/Downloads/Fielding.csv")
HOF <- read.csv("/Users/justincarter/Downloads/HallOfFame.csv")
batting <- batting[1:79248,]
batting <- batting %>% replace(is.na(.), 0)
batting <- batting %>% summarize(G = sum(G), R = sum(R), H = sum(H), XBH = sum(X2B)+sum(X3B),
HR = sum(HR), RBI = sum(RBI), SB = sum(SB), SO = sum(SO),
Ryear = max(yearID), .by = "playerID")
fielding <- filter(fielding, POS != "P")
fielding <- fielding %>% replace(is.na(.), 0)
fielding <- fielding %>% summarize(PO = sum(PO), A = sum(A), E = sum(E), .by = "playerID" )
position_players <- inner_join(batting, fielding, by = "playerID")
position_players$HOF <- 0
HOF <- filter(HOF, inducted == "Y" & category == "Player")
HOF <- semi_join(HOF, position_players, by = "playerID")
HOF <- HOF %>% arrange(HOF, playerID)
HOF_players <- semi_join(position_players, HOF, by = "playerID")
HOF_players <- HOF_players %>% arrange(HOF_players, playerID)
HOF_players$HOF = 1
HOF_players$years_to_induction <- (c(HOF$yearid) - c(HOF_players$Ryear))
non_HOF_Players <- anti_join(position_players, HOF, by = "playerID")
position_players <- full_join(HOF_players, non_HOF_Players)
## Joining with `by = join_by(playerID, G, R, H, XBH, HR, RBI, SB, SO, Ryear, PO,
## A, E, HOF)`
position_players$HOF <- as.factor(position_players$HOF)
HOF <- read.csv("/Users/justincarter/Downloads/HallOfFame.csv")
pitching <- read.csv("/Users/justincarter/Downloads/Pitching.csv")
pitching <- pitching[1:32898,]
pitching <- pitching %>% replace(is.na(.), 0)
pitching <- pitching %>% summarize(G = sum(G), W = sum(W), L = sum(L),
WHIP = ((sum(H) + sum(BB)) / (sum(IPouts) * 3)),
ERA = (sum(ER) / (sum(IPouts) / 3)) * 9,
SO = sum(SO), BB = sum(BB), HR = sum(HR), Ryear = max(yearID),
.by = "playerID")
pitching['ERA'][sapply(pitching['ERA'], is.infinite)] <- 10
pitching['ERA'][sapply(pitching['ERA'], is.nan)] <- 10
pitching['WHIP'][sapply(pitching['WHIP'], is.infinite)] <- 2
pitching['WHIP'][sapply(pitching['WHIP'], is.nan)] <- 2
HOF <- filter(HOF, inducted == "Y" & category == "Player")
HOF <- semi_join(HOF, pitching, by = "playerID")
HOF <- HOF %>% arrange(HOF, playerID)
HOF_pitchers <- semi_join(pitching, HOF, by = "playerID")
HOF_pitchers <- HOF_pitchers %>% arrange(HOF_pitchers, playerID)
HOF_pitchers$HOF = 1
HOF_pitchers$years_to_induction <- (c(HOF$yearid) - c(HOF_pitchers$Ryear))
HOF_pitchers <- filter(HOF_pitchers, G>100)
non_HOF_pitchers<- anti_join(pitching, HOF, by = "playerID")
non_HOF_pitchers$HOF = 0
pitchers <- full_join(HOF_pitchers, non_HOF_pitchers)
## Joining with `by = join_by(playerID, G, W, L, WHIP, ERA, SO, BB, HR, Ryear,
## HOF)`
pitchers$HOF <- as.factor(pitchers$HOF)
library(caTools)
library(randomForest)
## randomForest 4.7-1.1
## Type rfNews() to see new features/changes/bug fixes.
##
## Attaching package: 'randomForest'
## The following object is masked from 'package:dplyr':
##
## combine
library(caret)
## Loading required package: ggplot2
##
## Attaching package: 'ggplot2'
## The following object is masked from 'package:randomForest':
##
## margin
## Loading required package: lattice
set.seed(2002)
tuned_1 <- tuneRF(x = position_players[,c(2,3,4,5,6,7,8,9,11,12,13)], y = position_players[,14],
ntreeTry = 500, stepFactor = 1.5, improve = 1e-5 )
## mtry = 3 OOB error = 1.22%
## Searching left ...
## mtry = 2 OOB error = 1.31%
## -0.08108108 1e-05
## Searching right ...
## mtry = 4 OOB error = 1.2%
## 0.009009009 1e-05
## mtry = 6 OOB error = 1.24%
## -0.02727273 1e-05
tuned_2 <- tuneRF(x = pitchers[,c(2,3,4,5,6,7,8,9)], y = pitchers[,11],
ntreeTry = 500, stepFactor = 1.5, improve = 1e-5 )
## mtry = 2 OOB error = 0.52%
## Searching left ...
## Searching right ...
## mtry = 3 OOB error = 0.57%
## -0.08108108 1e-05
tuned_3 <- tuneRF(x = HOF_players[,c(2,3,4,5,6,7,8,9,11,12,13)], y = HOF_players[,15],
ntreeTry = 500, stepFactor = 1.5, improve = 1e-5 )
## mtry = 3 OOB error = 260.8404
## Searching left ...
## mtry = 2 OOB error = 264.7602
## -0.0150275 1e-05
## Searching right ...
## mtry = 4 OOB error = 260.3891
## 0.001730467 1e-05
## mtry = 6 OOB error = 265.0283
## -0.01781646 1e-05
tuned_4 <- tuneRF(x = HOF_pitchers[,c(2,3,4,5,6,7,8,9)], y = HOF_pitchers[,12],
ntreeTry = 500, stepFactor = 1.5, improve = 1e-5 )
## mtry = 2 OOB error = 292.4192
## Searching left ...
## Searching right ...
## mtry = 3 OOB error = 294.3751
## -0.006688694 1e-05
set.seed(2002)
data_split <- sample.split(position_players$G, SplitRatio = 0.75 )
train_1 <- filter(position_players, data_split == TRUE)
test_1 <- filter(position_players, data_split == FALSE)
train_1 <- as.data.frame(train_1)
test_1 <- as.data.frame(test_1)
RF_pPlayer <- randomForest(x = train_1[,c(2,3,4,5,6,7,8,9,11,12,13)], y = train_1[,14],
mtry = 4, ntree = 500)
plot(RF_pPlayer)
RF_pPlayer
##
## Call:
## randomForest(x = train_1[, c(2, 3, 4, 5, 6, 7, 8, 9, 11, 12, 13)], y = train_1[, 14], ntree = 500, mtry = 4)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 4
##
## OOB estimate of error rate: 1.46%
## Confusion matrix:
## 0 1 class.error
## 0 6946 24 0.003443329
## 1 80 89 0.473372781
predictions_1 <- predict(RF_pPlayer, newdata = test_1[,c(2,3,4,5,6,7,8,9,11,12,13)])
tab_1 <- confusionMatrix(data = predictions_1, reference = test_1[,14])
tab_1
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 1984 8
## 1 1 3
##
## Accuracy : 0.9955
## 95% CI : (0.9915, 0.9979)
## No Information Rate : 0.9945
## P-Value [Acc > NIR] : 0.3399
##
## Kappa : 0.3982
##
## Mcnemar's Test P-Value : 0.0455
##
## Sensitivity : 0.9995
## Specificity : 0.2727
## Pos Pred Value : 0.9960
## Neg Pred Value : 0.7500
## Prevalence : 0.9945
## Detection Rate : 0.9940
## Detection Prevalence : 0.9980
## Balanced Accuracy : 0.6361
##
## 'Positive' Class : 0
##
varImpPlot(RF_pPlayer, main = "Importance")
importance(RF_pPlayer)
## MeanDecreaseGini
## G 26.61221
## R 69.88409
## H 57.97436
## XBH 44.13089
## HR 15.83333
## RBI 36.81185
## SB 12.92611
## SO 17.29758
## PO 17.13721
## A 18.03437
## E 12.31427
set.seed(2002)
data_split_2 <- sample.split(HOF_players$playerID, SplitRatio = 0.75 )
train_2 <- filter(HOF_players, data_split_2 == TRUE)
test_2 <- filter(HOF_players, data_split_2 == FALSE)
train_2 <- as.data.frame(train_2)
test_2 <- as.data.frame(test_2)
RF_HOF_Pplayers <- randomForest(formula = years_to_induction ~ G + R + H + XBH + HR + RBI + SB
+ SO + PO + E + A, data = train_2, mtry = 2, ntree = 500)
plot(RF_HOF_Pplayers)
RF_HOF_Pplayers
##
## Call:
## randomForest(formula = years_to_induction ~ G + R + H + XBH + HR + RBI + SB + SO + PO + E + A, data = train_2, mtry = 2, ntree = 500)
## Type of random forest: regression
## Number of trees: 500
## No. of variables tried at each split: 2
##
## Mean of squared residuals: 293.1835
## % Var explained: 40.17
varImpPlot(RF_HOF_Pplayers, main = "Importance")
importance(RF_HOF_Pplayers)
## IncNodePurity
## G 6298.900
## R 3352.168
## H 4192.191
## XBH 3659.027
## HR 7640.961
## RBI 4908.941
## SB 3400.154
## SO 7018.880
## PO 5624.812
## E 10487.636
## A 3962.662
set.seed(2002)
data_split_3 <- sample.split(pitchers$playerID, SplitRatio = 0.75 )
train_3 <- filter(pitchers, data_split_3 == TRUE)
test_3<- filter(pitchers, data_split_3 == FALSE)
train_3 <- as.data.frame(train_3)
test_3 <- as.data.frame(test_3)
RF_pitchers <- randomForest(x = train_3[,c(2,3,4,5,6,7,8,9)], y = train_3[,11],
mtry = 4, ntree = 500)
plot(RF_pitchers)
RF_pitchers
##
## Call:
## randomForest(x = train_3[, c(2, 3, 4, 5, 6, 7, 8, 9)], y = train_3[, 11], ntree = 500, mtry = 4)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 4
##
## OOB estimate of error rate: 0.55%
## Confusion matrix:
## 0 1 class.error
## 0 5243 9 0.001713633
## 1 20 31 0.392156863
predictions_2 <- predict(RF_pitchers, newdata = test_3[,c(2,3,4,5,6,7,8,9)])
tab_2 <- confusionMatrix(data = predictions_2, reference = test_3[,11])
tab_2
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 1743 5
## 1 5 15
##
## Accuracy : 0.9943
## 95% CI : (0.9896, 0.9973)
## No Information Rate : 0.9887
## P-Value [Acc > NIR] : 0.01048
##
## Kappa : 0.7471
##
## Mcnemar's Test P-Value : 1.00000
##
## Sensitivity : 0.9971
## Specificity : 0.7500
## Pos Pred Value : 0.9971
## Neg Pred Value : 0.7500
## Prevalence : 0.9887
## Detection Rate : 0.9859
## Detection Prevalence : 0.9887
## Balanced Accuracy : 0.8736
##
## 'Positive' Class : 0
##
varImpPlot(RF_pitchers, main = "Importance")
importance(RF_pitchers)
## MeanDecreaseGini
## G 7.325837
## W 33.527092
## L 7.696084
## WHIP 9.568001
## ERA 8.420895
## SO 20.828467
## BB 7.082479
## HR 5.053687
set.seed(2002)
data_split_4 <- sample.split(HOF_pitchers$playerID, SplitRatio = 0.75 )
train_4 <- filter(HOF_pitchers, data_split_4 == TRUE)
test_4 <- filter(HOF_pitchers, data_split_4 == FALSE)
train_4 <- as.data.frame(train_4)
test_4 <- as.data.frame(test_4)
RF_HOF_pitchers <- randomForest(formula = years_to_induction ~ G + W + L + WHIP +
ERA + SO + BB + HR, data = HOF_pitchers, mtry = 2, ntree = 500)
plot(RF_HOF_pitchers)
RF_HOF_pitchers
##
## Call:
## randomForest(formula = years_to_induction ~ G + W + L + WHIP + ERA + SO + BB + HR, data = HOF_pitchers, mtry = 2, ntree = 500)
## Type of random forest: regression
## Number of trees: 500
## No. of variables tried at each split: 2
##
## Mean of squared residuals: 287.3263
## % Var explained: 38
varImpPlot(RF_HOF_pitchers, main = "Importance")
importance(RF_HOF_pitchers)
## IncNodePurity
## G 2485.978
## W 3449.482
## L 3590.861
## WHIP 2913.879
## ERA 3753.931
## SO 4146.651
## BB 3225.618
## HR 6441.786
log_Pplayers <- glm(HOF ~ G + R + H + XBH + HR + RBI + SB
+ SO + PO + E + A, data = train_1, family = "binomial")
summary(log_Pplayers)
##
## Call:
## glm(formula = HOF ~ G + R + H + XBH + HR + RBI + SB + SO + PO +
## E + A, family = "binomial", data = train_1)
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -6.285e+00 2.678e-01 -23.471 < 2e-16 ***
## G -1.526e-03 7.579e-04 -2.014 0.044032 *
## R 4.425e-03 1.124e-03 3.938 8.20e-05 ***
## H 8.658e-04 9.757e-04 0.887 0.374906
## XBH -1.421e-03 2.554e-03 -0.557 0.577800
## HR 5.559e-04 3.302e-03 0.168 0.866302
## RBI 3.933e-03 1.277e-03 3.080 0.002069 **
## SB -3.297e-04 1.061e-03 -0.311 0.756023
## SO -1.631e-03 5.902e-04 -2.764 0.005712 **
## PO -5.039e-05 2.938e-05 -1.715 0.086382 .
## E -3.693e-03 1.105e-03 -3.344 0.000827 ***
## A 3.070e-04 7.842e-05 3.915 9.05e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 1599.25 on 7138 degrees of freedom
## Residual deviance: 707.88 on 7127 degrees of freedom
## AIC: 731.88
##
## Number of Fisher Scoring iterations: 8
predictions_3 <- rep(0, length(test_1[,1]))
probs_1 <- predict(log_Pplayers, newdata = test_1[,c(2,3,4,5,6,7,8,9,11,12,13)], type = "response")
predictions_3[probs_1 > 0.5] <- 1
predictions_3 <- as.factor(predictions_3)
tab_3 <- confusionMatrix(data = predictions_3, reference = test_1[,14])
tab_3
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 1983 10
## 1 2 1
##
## Accuracy : 0.994
## 95% CI : (0.9895, 0.9969)
## No Information Rate : 0.9945
## P-Value [Acc > NIR] : 0.68900
##
## Kappa : 0.1408
##
## Mcnemar's Test P-Value : 0.04331
##
## Sensitivity : 0.99899
## Specificity : 0.09091
## Pos Pred Value : 0.99498
## Neg Pred Value : 0.33333
## Prevalence : 0.99449
## Detection Rate : 0.99349
## Detection Prevalence : 0.99850
## Balanced Accuracy : 0.54495
##
## 'Positive' Class : 0
##
lm_HOF_Pplayers <- lm(years_to_induction ~ G + R + H + XBH + HR + RBI + SB
+ SO + PO + E + A, data = train_2)
summary(lm_HOF_Pplayers)
##
## Call:
## lm(formula = years_to_induction ~ G + R + H + XBH + HR + RBI +
## SB + SO + PO + E + A, data = train_2)
##
## Residuals:
## Min 1Q Median 3Q Max
## -31.135 -9.486 -1.926 7.496 46.162
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 42.6903821 4.5397511 9.404 3.73e-16 ***
## G -0.0061555 0.0093548 -0.658 0.51177
## R 0.0180863 0.0135624 1.334 0.18481
## H -0.0113898 0.0114601 -0.994 0.32224
## XBH -0.0197346 0.0270371 -0.730 0.46683
## HR -0.0651313 0.0337321 -1.931 0.05580 .
## RBI 0.0126175 0.0134200 0.940 0.34896
## SB -0.0096986 0.0113123 -0.857 0.39292
## SO 0.0023027 0.0066239 0.348 0.72871
## PO -0.0005059 0.0004017 -1.260 0.21023
## E 0.0560952 0.0130605 4.295 3.51e-05 ***
## A -0.0023236 0.0008421 -2.760 0.00667 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 16.55 on 123 degrees of freedom
## Multiple R-squared: 0.491, Adjusted R-squared: 0.4455
## F-statistic: 10.79 on 11 and 123 DF, p-value: 1.055e-13
log_pitchers <- glm(HOF ~ G + W + L + WHIP + ERA + SO + BB + HR, data = train_3,
family = "binomial")
## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
summary(log_pitchers)
##
## Call:
## glm(formula = HOF ~ G + W + L + WHIP + ERA + SO + BB + HR, family = "binomial",
## data = train_3)
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -4.724e+00 1.683e+00 -2.806 0.005018 **
## G 5.651e-03 1.405e-03 4.023 5.76e-05 ***
## W 4.461e-02 8.168e-03 5.462 4.70e-08 ***
## L -3.677e-02 1.007e-02 -3.651 0.000261 ***
## WHIP -1.225e+01 1.945e+01 -0.630 0.528739
## ERA -1.050e+00 6.471e-01 -1.623 0.104500
## SO 1.382e-03 6.255e-04 2.209 0.027186 *
## BB 8.468e-04 1.031e-03 0.821 0.411388
## HR -4.399e-03 3.422e-03 -1.286 0.198553
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 575.22 on 5302 degrees of freedom
## Residual deviance: 156.57 on 5294 degrees of freedom
## AIC: 174.57
##
## Number of Fisher Scoring iterations: 15
predictions_4 <- rep(0, length(test_3[,1]))
probs_2 <- predict(log_pitchers, newdata = test_3[,c(2,3,4,5,6,7,8,9)])
predictions_4[probs_2 > 0.5] <- 1
predictions_4 <- as.factor(predictions_4)
tab_4 <- confusionMatrix(data = predictions_4, reference = test_3[,11])
tab_4
## Confusion Matrix and Statistics
##
## Reference
## Prediction 0 1
## 0 1746 8
## 1 2 12
##
## Accuracy : 0.9943
## 95% CI : (0.9896, 0.9973)
## No Information Rate : 0.9887
## P-Value [Acc > NIR] : 0.01048
##
## Kappa : 0.7031
##
## Mcnemar's Test P-Value : 0.11385
##
## Sensitivity : 0.9989
## Specificity : 0.6000
## Pos Pred Value : 0.9954
## Neg Pred Value : 0.8571
## Prevalence : 0.9887
## Detection Rate : 0.9876
## Detection Prevalence : 0.9921
## Balanced Accuracy : 0.7994
##
## 'Positive' Class : 0
##
lm_HOF_pitchers <- lm(years_to_induction~ G + W + L + WHIP + ERA + SO + BB + HR,
data = train_4,)
summary(lm_HOF_pitchers)
##
## Call:
## lm(formula = years_to_induction ~ G + W + L + WHIP + ERA + SO +
## BB + HR, data = train_4)
##
## Residuals:
## Min 1Q Median 3Q Max
## -34.450 -7.331 -0.654 6.936 38.181
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 7.269e+01 6.632e+01 1.096 0.279014
## G -2.980e-02 1.530e-02 -1.948 0.057850 .
## W -9.125e-02 5.867e-02 -1.555 0.127042
## L 3.142e-01 8.327e-02 3.773 0.000478 ***
## WHIP -1.064e+02 7.904e+02 -0.135 0.893575
## ERA -9.852e+00 1.626e+01 -0.606 0.547792
## SO -7.779e-03 6.477e-03 -1.201 0.236172
## BB 1.012e-02 1.560e-02 0.649 0.519770
## HR -8.063e-02 3.547e-02 -2.273 0.027963 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 15.34 on 44 degrees of freedom
## Multiple R-squared: 0.5499, Adjusted R-squared: 0.4681
## F-statistic: 6.72 on 8 and 44 DF, p-value: 1.004e-05