def calculate_metric_value(knn_model, knn_df):
result = knn_model.transform(knn_df)
# map predicted result string types to numeric type
indexer = StringIndexer()
indexer.setInputCol('predict')
indexer.setOutputCol('predictIndexer')
index_model = indexer.fit(result)
index_result = index_model.transform(result)
# generate label column
dataset = index_result.withColumn('label', expr("predictIndexer"))
evaluator = MulticlassClassificationEvaluator(
predictionCol="predictIndexer")
# calculate metric
metric_dict["f1"] = (evaluator.evaluate(dataset,
{evaluator.metricName: "f1"}))
metric_dict["accuracy"] = (evaluator.evaluate(
dataset, {evaluator.metricName: "accuracy"}))
metric_dict["precision"] = (evaluator.evaluate(
dataset, {evaluator.metricName: "weightedPrecision"}))
metric_dict["recall"] = (evaluator.evaluate(
dataset, {evaluator.metricName: "weightedRecall"}))
