Skip to contents

Predicting Infectious Diseases via Environment and Climate (PRIDE-C)

The goal of PRIDE-C is to provide a standardized API functionality for forecasting infectious diseases from DHIS2 data.

Installation

You can install the development version of PRIDEC from GitHub with:

# install.packages("devtools")
devtools::install_github("Pivot-Madagascar/PRIDEC-package")

Example

TThe PRIDE-C forecasting approach follows the following steps:

  1. Data processing
  2. Model tuning and training
  3. Forecasting

The example below follows the steps for fitting a Random Forest model using ranger from simulated model data.

library(PRIDEC)

data(demo_malaria)

#preprocess data set
data_clean <- prep_data(raw_data = demo_malaria,
                        y_var = "n_case",
                        lagged_vars =  c("rain_mm", "temp_c"),
                        scaled_vars = NULL,
                        graph_poly = NULL)
#> Registered S3 method overwritten by 'quantmod':
#>   method            from
#>   as.zoo.data.frame zoo
#create a cv_fold of assessment (historical) and analysis (forecast) data
cv_set <- split_cv_rolling(data_clean$data_prep, 
                           month_analysis = 48, 
                           month_assess=3)[[20]]

#fit the model and forecast
rf_fit <- fit_ranger(cv_set, 
                     y_var = "n_case",
                     id_vars = c("orgUnit", "date"),
                     pred_vars = c("rain_mm", "temp_c", "month_season", "orgUnit"))

#evaluate model fit
eval_performance(rf_fit)
#> Registered S3 method overwritten by 'scoringutils':
#>   method         from    
#>   print.forecast forecast
#> # A tibble: 2 × 10
#>   dataset    wis   mae med_ae mean_ae_log  wape dispersion sp_rho prop_over
#>   <chr>    <dbl> <dbl>  <dbl>       <dbl> <dbl>      <dbl>  <dbl>     <dbl>
#> 1 analysis  5.86  6.70  0.155      0.0137 0.155       5.07  0.963    0.207 
#> 2 assess   24.4  33.3  32.7        1.09   0.611       2.36  0.821    0.0317
#> # ℹ 1 more variable: prop_under <dbl>
plot_predictions(rf_fit[rf_fit$orgUnit %in% sample(rf_fit$orgUnit,1),])

Contribute to PRIDE-C

As an open-source package, we welcome all contributions. Please feel free to file an issue or contact the developer (@mvevans89).

Funding

The development and maintenance of this package is funded by a Wellcome Trust Digital Technology Development Award.