Water Potability
Ridho Hilmansyah Botutihe
2022-06-03
Objective
Access to safe drinking-water is essential to health, a basic human right and a component of effective policy for health protection. This is important as a health and development issue at a national, regional and local level. In some regions, it has been shown that investments in water supply and sanitation can yield a net economic benefit, since the reductions in adverse health effects and health care costs outweigh the costs of undertaking the interventions.
Data Preparation
#Library loading
library(dplyr)
library(e1071)
library(rsample)
library(caret)
library(ROCR)
library(partykit)
library(rsample)
library(randomForest)
library(rpart)#Import csv data to data frame
water <- read.csv("water_potability.csv")
waterData Wrangling
#Inspecting data
glimpse(water)#> Rows: 3,276
#> Columns: 10
#> $ ph <dbl> NA, 3.716080, 8.099124, 8.316766, 9.092223, 5.584087, ~
#> $ Hardness <dbl> 204.8905, 129.4229, 224.2363, 214.3734, 181.1015, 188.~
#> $ Solids <dbl> 20791.32, 18630.06, 19909.54, 22018.42, 17978.99, 2874~
#> $ Chloramines <dbl> 7.300212, 6.635246, 9.275884, 8.059332, 6.546600, 7.54~
#> $ Sulfate <dbl> 368.5164, NA, NA, 356.8861, 310.1357, 326.6784, 393.66~
#> $ Conductivity <dbl> 564.3087, 592.8854, 418.6062, 363.2665, 398.4108, 280.~
#> $ Organic_carbon <dbl> 10.379783, 15.180013, 16.868637, 18.436524, 11.558279,~
#> $ Trihalomethanes <dbl> 86.99097, 56.32908, 66.42009, 100.34167, 31.99799, 54.~
#> $ Turbidity <dbl> 2.963135, 4.500656, 3.055934, 4.628771, 4.075075, 2.55~
#> $ Potability <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ~#Checking Missing Value
colSums(is.na(water))#> ph Hardness Solids Chloramines Sulfate
#> 491 0 0 0 781
#> Conductivity Organic_carbon Trihalomethanes Turbidity Potability
#> 0 0 162 0 0#Fill NA value with mean and change potability column to factor
water_clean <- water %>%
group_by(Potability) %>%
mutate_at(vars(-c("Potability")),~ifelse(is.na(.), mean(., na.rm = TRUE), .)) %>%
mutate(Potability = as.factor(Potability),
ph = as.numeric(ph),
Hardness = as.numeric(Hardness),
Solids = as.numeric(Solids),
Chloramines = as.numeric(Chloramines),
Sulfate = as.numeric(Sulfate),
Conductivity = as.numeric(Conductivity),
Organic_carbon = as.numeric(Organic_carbon),
Trihalomethanes = as.numeric(Trihalomethanes),
Turbidity = as.numeric(Turbidity))
water_cleanExploratory Data Analysis
#Checking target variable proportion
prop.table(table(water_clean$Potability))#>
#> 0 1
#> 0.6098901 0.3901099We can assume that the target variable is balanced, with 0 (Not potable) at 60%, and 1 (potable) at 40%
Cross-Validation
#Splitting data to data train (80%) and data test (20%)
RNGkind(sample.kind = "Rounding")
set.seed(1234)
index <- sample(x = nrow(water_clean),
size = nrow(water_clean)*0.8)
water_train <- water_clean[index, ]
water_test <- water_clean[-index, ]Naive-Bayes Model
Model Fitting
#Building a naive bayes model
water_naive <- naiveBayes(Potability ~ ., data = water_train)
water_naive#>
#> Naive Bayes Classifier for Discrete Predictors
#>
#> Call:
#> naiveBayes.default(x = X, y = Y, laplace = laplace)
#>
#> A-priori probabilities:
#> Y
#> 0 1
#> 0.6064885 0.3935115
#>
#> Conditional probabilities:
#> ph
#> Y [,1] [,2]
#> 0 7.075419 1.544355
#> 1 7.059120 1.353061
#>
#> Hardness
#> Y [,1] [,2]
#> 0 196.9079 30.89728
#> 1 195.9732 35.60567
#>
#> Solids
#> Y [,1] [,2]
#> 0 21842.25 8519.572
#> 1 22320.76 8919.007
#>
#> Chloramines
#> Y [,1] [,2]
#> 0 7.056088 1.478515
#> 1 7.188880 1.713786
#>
#> Sulfate
#> Y [,1] [,2]
#> 0 334.6017 31.67072
#> 1 331.8996 40.85486
#>
#> Conductivity
#> Y [,1] [,2]
#> 0 426.1444 78.47286
#> 1 426.0730 81.50378
#>
#> Organic_carbon
#> Y [,1] [,2]
#> 0 14.39037 3.318138
#> 1 14.18224 3.197548
#>
#> Trihalomethanes
#> Y [,1] [,2]
#> 0 66.50815 15.57059
#> 1 66.49571 15.77251
#>
#> Turbidity
#> Y [,1] [,2]
#> 0 3.964457 0.7797357
#> 1 3.958181 0.7827792Model Evaluation
# Predict model with data test
water_test$pred <- predict(water_naive, newdata = water_test, type = "class" )
head(water_test)# confusion Matrix
confusionMatrix(data = water_test$pred, reference = water_test$Potability,
positive = "1")#> Confusion Matrix and Statistics
#>
#> Reference
#> Prediction 0 1
#> 0 352 181
#> 1 57 66
#>
#> Accuracy : 0.6372
#> 95% CI : (0.5991, 0.6741)
#> No Information Rate : 0.6235
#> P-Value [Acc > NIR] : 0.2473
#>
#> Kappa : 0.142
#>
#> Mcnemar's Test P-Value : 1.55e-15
#>
#> Sensitivity : 0.2672
#> Specificity : 0.8606
#> Pos Pred Value : 0.5366
#> Neg Pred Value : 0.6604
#> Prevalence : 0.3765
#> Detection Rate : 0.1006
#> Detection Prevalence : 0.1875
#> Balanced Accuracy : 0.5639
#>
#> 'Positive' Class : 1
#> Decision Tree
water_tree <- rpart(Potability ~ ., method = "class", data = water_train)
water_tree_preds <- predict(water_tree, water_test, type = "class")
confusionMatrix(water_tree_preds, water_test$Potability, positive='1')#> Confusion Matrix and Statistics
#>
#> Reference
#> Prediction 0 1
#> 0 360 104
#> 1 49 143
#>
#> Accuracy : 0.7668
#> 95% CI : (0.7325, 0.7986)
#> No Information Rate : 0.6235
#> P-Value [Acc > NIR] : 3.153e-15
#>
#> Kappa : 0.4803
#>
#> Mcnemar's Test P-Value : 1.268e-05
#>
#> Sensitivity : 0.5789
#> Specificity : 0.8802
#> Pos Pred Value : 0.7448
#> Neg Pred Value : 0.7759
#> Prevalence : 0.3765
#> Detection Rate : 0.2180
#> Detection Prevalence : 0.2927
#> Balanced Accuracy : 0.7296
#>
#> 'Positive' Class : 1
#> Random Forest
water_forest <- randomForest(Potability ~ ., data= water_train, ntree= 1000)
water_forest_preds <- predict(water_forest, water_test, type = "class")
confusionMatrix(water_forest_preds, water_test$Potability, positive='1')#> Confusion Matrix and Statistics
#>
#> Reference
#> Prediction 0 1
#> 0 354 82
#> 1 55 165
#>
#> Accuracy : 0.7912
#> 95% CI : (0.758, 0.8217)
#> No Information Rate : 0.6235
#> P-Value [Acc > NIR] : < 2e-16
#>
#> Kappa : 0.5453
#>
#> Mcnemar's Test P-Value : 0.02633
#>
#> Sensitivity : 0.6680
#> Specificity : 0.8655
#> Pos Pred Value : 0.7500
#> Neg Pred Value : 0.8119
#> Prevalence : 0.3765
#> Detection Rate : 0.2515
#> Detection Prevalence : 0.3354
#> Balanced Accuracy : 0.7668
#>
#> 'Positive' Class : 1
#> Conclusion
Predicting not drinkable water as drinkable (false positive) is in my opinion the most crucial thing to avoid. Therefore, specificity is the most important measure. Because, a high specificity means many true negatives and few false positives. Comparing models above, All models perform almost the same, with 85% plus in specificity metrics