Air Quality Dataset

Description

These datasets contain air quality measurements for training and testing purposes. They include various air pollutant concentrations and meteorological variables measured at different locations and times.

Usage

data("Air_quality_training")
data("Air_quality_testing")

Format

Both datasets are data frames with 8760 rows and 12 variables:

Date

Date and time of measurement (POSIXct format)

PM2.5

Particulate matter with diameter less than 2.5 micrometers (\mu g/m^3)

PM10

Particulate matter with diameter less than 10 micrometers (\mu g/m^3)

SO2

Sulfur dioxide concentration (\mu g/m^3)

NO2

Nitrogen dioxide concentration (\mu g/m^3)

CO

Carbon monoxide concentration (\mu g/m^3)

O3

Ozone concentration (\mu g/m^3)

TEMP

Temperature (\textdegree C)

PRES

Atmospheric pressure (hPa)

DEWP

Dew point temperature (\textdegree C)

RAIN

Precipitation amount (mm)

WSPM

Wind speed (m/s)

Details

Dataset Differences:

• Air_quality_training: Used for training SCA and SCE models

• Air_quality_testing: Used for testing trained models

Variable Descriptions:

• PM2.5, PM10: Particulate matter concentrations, important indicators of air quality

• SO2, NO2, CO, O3: Major air pollutants regulated by environmental agencies

• TEMP, PRES, DEWP: Meteorological variables affecting air quality

• RAIN, WSPM: Weather conditions that influence pollutant dispersion

Source

Air quality monitoring stations

Plot Recursive Feature Elimination Results

Description

Plot Recursive Feature Elimination results.

Usage

Plot_RFE(rfe_result, 
         main = "OOB Validation and Testing R2 vs Number of Predictors", 
         col_validation = "blue", 
         col_testing = "red", 
         pch = 16, 
         lwd = 2, 
         cex = 1.2, 
         legend_pos = "bottomleft", 
         ...)

Arguments

Value

Plot showing validation and testing R2 vs number of predictors.

See Also

RFE_SCE

Recursive Feature Elimination for SCE Models

Description

Recursive Feature Elimination for SCE models to identify the most important predictors.

Usage

RFE_SCE(Training_data, Testing_data, Predictors, Predictant, Nmin, Ntree, 
        alpha = 0.05, resolution = 1000, step = 1, verbose = TRUE, 
        parallel = TRUE)

Arguments

Value

RFE results with performance metrics and importance scores.

See Also

Plot_RFE, SCE, importance

Stepwise Cluster Analysis (SCA)

Description

Builds a single Stepwise Cluster Analysis (SCA) tree model that recursively partitions the data space based on Wilks' Lambda statistic.

Usage

SCA(Training_data, X, Y, Nmin, alpha = 0.05, resolution = 1000, verbose = FALSE)

Arguments

Value

An S3 object of class "SCA" containing the tree model.

See Also

SCE, predict, importance, evaluate

Examples

  # Load example data
  data(Streamflow_training_10var)
  data(Streamflow_testing_10var)
  
  # Define variables
  Predictors <- c("Prcp","SRad","Tmax","Tmin","VP","smlt","swvl1","swvl2","swvl3","swvl4")
  Predictants <- c("Flow")
  
  # Build SCA model
  sca_model <- SCA(
    Training_data = Streamflow_training_10var,
    X = Predictors,
    Y = Predictants,
    Nmin = 5,
    alpha = 0.05,
    resolution = 1000
  )
  
  # Use S3 methods
  print(sca_model)
  summary(sca_model)
  sca_predictions <- predict(sca_model, Streamflow_testing_10var)
  sca_importance <- importance(sca_model)
  sca_evaluation <- evaluate(sca_model, Streamflow_testing_10var, Streamflow_training_10var)

Stepwise Clustered Ensemble (SCE)

Description

Builds a Stepwise Clustered Ensemble (SCE) model, which is an ensemble of SCA trees built using bootstrap samples and random feature selection, providing improved prediction accuracy and robustness.

Usage

SCE(Training_data, X, Y, mfeature, Nmin, Ntree, alpha = 0.05, 
    resolution = 1000, verbose = FALSE, parallel = TRUE)

Arguments

Value

An S3 object of class "SCE" containing the ensemble model.

See Also

SCA, predict, importance, evaluate

Examples

  # Load example data
  data(Streamflow_training_10var)
  data(Streamflow_testing_10var)
  
  # Define variables
  Predictors <- c("Prcp","SRad","Tmax","Tmin","VP","smlt","swvl1","swvl2","swvl3","swvl4")
  Predictants <- c("Flow")
  
  # Build SCE model
  sce_model <- SCE(
    Training_data = Streamflow_training_10var,
    X = Predictors,
    Y = Predictants,
    mfeature = round(0.5 * length(Predictors)),
    Nmin = 5,
    Ntree = 48,
    alpha = 0.05,
    resolution = 1000,
    parallel = FALSE
  )
  
  # Use S3 methods
  print(sce_model)
  summary(sce_model)
  sce_predictions <- predict(sce_model, Streamflow_testing_10var)
  sce_importance <- importance(sce_model)
  sce_evaluation <- evaluate(sce_model, Streamflow_testing_10var, Streamflow_training_10var)

Streamflow Dataset

Description

These datasets contain streamflow and related environmental variables for training and testing purposes. They are used in examples to demonstrate the SCE package functionality with different levels of complexity.

Usage

data("Streamflow_training_10var")
data("Streamflow_training_22var")
data("Streamflow_testing_10var")
data("Streamflow_testing_22var")

Format

Streamflow_training_10var: Basic environmental variables (12 columns):

Date

Date and time of measurement

Prcp

Monthly mean daily precipitation (mm)

SRad

Monthly mean daily solar radiation (W/m^2)

Tmax

Monthly mean daily maximum temperature (°C)

Tmin

Monthly mean daily minimum temperature (°C)

VP

Monthly mean daily vapor pressure (Pa)

smlt

Monthly snowmelt (m)

swvl1

Soil water content layer 1 (m^3/m^3)

swvl2

Soil water content layer 2 (m^3/m^3)

swvl3

Soil water content layer 3 (m^3/m^3)

swvl4

Soil water content layer 4 (m^3/m^3)

Flow

Monthly mean daily streamflow (cfs)

Streamflow_training_22var: Extended variables with climate indices (24 columns):

Flow

Streamflow measurements

IPO

Interdecadal Pacific Oscillation

IPO_lag1

IPO with 1-month lag

IPO_lag2

IPO with 2-month lag

Nino3.4

Nino 3.4 index

Nino3.4_lag1

Nino 3.4 with 1-month lag

Nino3.4_lag2

Nino 3.4 with 2-month lag

PDO

Pacific Decadal Oscillation

PDO_lag1

PDO with 1-month lag

PDO_lag2

PDO with 2-month lag

PNA

Pacific North American pattern

PNA_lag1

PNA with 1-month lag

PNA_lag2

PNA with 2-month lag

Precipitation

Monthly precipitation

Precipitation_2Mon

2-month precipitation

Radiation

Solar radiation

Radiation_2Mon

2-month solar radiation

Tmax

Maximum temperature

Tmax_2Mon

2-month maximum temperature

Tmin

Minimum temperature

Tmin_2Mon

2-month minimum temperature

VP

Vapor pressure

VP_2Mon

2-month vapor pressure

Testing datasets: Same structure as corresponding training datasets.

Details

Dataset Structure:

• 10var datasets: Basic environmental variables (12 columns)

• 22var datasets: Extended variables with climate indices (24 columns)

• Training datasets: Used for model building

• Testing datasets: Used for model evaluation

Climate Indices: IPO (Interdecadal Pacific Oscillation), Nino3.4 (El Niño), PDO (Pacific Decadal Oscillation), PNA (Pacific North American pattern)

Data Sources: ERA5 Land, Daymet, USGS, and climate indices databases

Source

Environmental monitoring stations, climate indices databases, ERA5 Land, Daymet, and USGS

Evaluate SCE and SCA Model Performance

Description

Evaluate model performance for SCE or SCA models.

Usage

## S3 method for class 'SCE'
evaluate(object, Testing_data, Training_data, digits = 3, ...)
## S3 method for class 'SCA'
evaluate(object, Testing_data, Training_data, digits = 3, ...)

Arguments

Value

Model performance metrics.

See Also

SCE, SCA, predict

Variable Importance for SCE and SCA Models

Description

Calculate variable importance for SCE or SCA models.

Usage

## S3 method for class 'SCE'
importance(object, OOB_weight = TRUE, digits = 2, ...)
## S3 method for class 'SCA'
importance(object, digits = 2, ...)

Arguments

Value

Variable importance rankings. For convenience, relative importance values are rounded to digits decimal places.

See Also

SCE, SCA, RFE_SCE

Predict Using SCE and SCA Models

Description

Make predictions on new data using SCE or SCA models.

Usage

## S3 method for class 'SCE'
predict(object, newdata, ...)
## S3 method for class 'SCA'
predict(object, newdata, ...)

Arguments

Value

Predictions for the new data.

See Also

SCE, SCA, evaluate

Print SCE and SCA Model Objects

Description

Print information about SCE or SCA model objects.

Usage

## S3 method for class 'SCE'
print(x, ...)
## S3 method for class 'SCA'
print(x, ...)

Arguments

Details

For SCE objects, prints ensemble information including number of trees, parameters, predictors, predictants, and OOB performance metrics.

For SCA objects, prints tree structure information including total nodes, leaf nodes, cutting/merging actions, and variable names.

Value

Prints model information and returns the object invisibly.

See Also

SCE, SCA, summary

Summary methods for SCE and SCA models

Description

Provide concise summaries of model structure and performance for SCE and SCA objects.

Usage

## S3 method for class 'SCE'
summary(object, ...)
## S3 method for class 'SCA'
summary(object, ...)

Arguments

Details

For summary.SCE, the method prints ensemble configuration, out-of-bag (OOB) performance statistics, tree structure information, and tree weight distribution.\ For summary.SCA, the method prints tree structure information and variable summaries for the single SCA tree.

Value

Invisibly returns the input object after printing the summary.

See Also

SCE, SCA, print, importance, evaluate