AR_example_workflow_Eunomia.Rmd

---
title: "An example for Association Rule Mining using the Eunomia package"
date: "11/12/2020"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(results = 'hide')
library(DatabaseConnector)
library(SqlRender)
library(Eunomia)
library(FeatureExtraction)
library(AssociationRuleMining)
devtools::load_all()
```

### Connect to the database 

```{r}
### Define database parameters
cdmdatabaseschema = "main"
resultsdatabaseschema = "main"
arm_inputFile = "testing.txt"
arm_outputFile = "testingResults.txt"

connectionDetails <- Eunomia::getEunomiaConnectionDetails()
connection <- connect(connectionDetails)
#on.exit(DatabaseConnector::disconnect(connection)) #Close db connection on error or exit

```

#### Define cohort ####

There are two options for defining a cohort:
- 1) Define the cohort in ATLAS and export the SQL file
- 2) define it locally within R

```{r}
cohort <- "SELECT person_id AS subject_id,
  condition_start_date AS cohort_start_date
INTO #diagnoses
FROM @cdm.condition_occurrence
WHERE condition_concept_id IN (
    SELECT descendant_concept_id
    FROM @cdm.concept_ancestor
    WHERE ancestor_concept_id = 4329847 -- Myocardial infarction
)
  AND condition_concept_id NOT IN (
    SELECT descendant_concept_id
    FROM @cdm.concept_ancestor
    WHERE ancestor_concept_id = 314666 -- Old myocardial infarction
);
INSERT INTO @cdm.cohort (subject_id, cohort_start_date, cohort_definition_id)
SELECT subject_id,
  cohort_start_date,
  CAST (1 AS INT) AS cohort_definition_id
FROM #diagnoses
INNER JOIN @cdm.visit_occurrence
  ON subject_id = person_id
    AND cohort_start_date >= visit_start_date
    AND cohort_start_date <= visit_end_date
WHERE visit_concept_id IN (9201, 9203, 262); -- Inpatient or ER;"

renderTranslateExecuteSql(connection, cohort, cdm = cdmdatabaseschema)

sql <- "ALTER TABLE #diagnoses ADD cohort_definition_id INT NOT NULL DEFAULT(1)"

# Execute the script to receive the data
renderTranslateExecuteSql(connection, sql)

querySql(connection, "SELECT count(*) FROM diagnoses;")
```

### Get the data and close the connection

```{r}
#### Feature Extraction ####
covariateSettings <- createCovariateSettings(useConditionOccurrenceAnyTimePrior = TRUE, 
                                             #includedCovariateIds = c(), 
                                             #includedCovariateConceptIds = c()
                                             )


covariateData_eunomia <- getDbCovariateData(connection = connection, cdmDatabaseSchema = cdmdatabaseschema, cohortDatabaseSchema = resultsDatabaseSchema, cohortTable = "diagnoses", rowIdField = "subject_id", covariateSettings = covariateSettings, cohortTableIsTemp = TRUE)

disconnect(connection)
```

```{r}
getInputFileForAssociationRules(covariateDataObject = covariateData_eunomia, fileToSave = arm_inputFile)
```

## Run Apriori

```{r}
apriori_associationSets <- runAssociationRules(algorithm = "Apriori", 
                                       inputFile = arm_inputFile, 
                                       outputFile = arm_outputFile,
                                       minsup = 0.5 )
head(apriori_associationSets)
```


## Run Eclat

```{r}
eclat_associationSets <- runAssociationRules(algorithm = "Eclat", 
                                       inputFile = arm_inputFile, 
                                       outputFile = arm_outputFile,
                                       minsup = 0.5 )
head(eclat_associationSets)
```

## Run FP-Growth 

```{r}
fpgrowth_associationSets <- runAssociationRules(algorithm = "FP-Growth", 
                                       inputFile = arm_inputFile, 
                                       outputFile = arm_outputFile,
                                       minsup = 0.5 )
head(fpgrowth_associationSets)
```

## Run Relim

```{r}
relim_associationSets <- runAssociationRules(algorithm = "Relim", 
                                       inputFile = arm_inputFile, 
                                       outputFile = arm_outputFile,
                                       minsup = 0.5 )
head(relim_associationSets)
```