Merge pull request #38 from McGill-MMA-EnterpriseAnalytics/aladelca-p…

…atch-5

predict

09_model_pipeline/predict/src/main_predict.py

@@ -0,0 +1,35 @@
+import json
+import pandas as pd
+from utils import *
+from predict import Predict as pr
+def lambda_handler(event, context):
+    bucket_name = 'enterprise-data-science'
+    path = 'training_files/trained_model.pickle'
+    #client = boto3.client('s3')
+
+    data = json.loads(json.dumps(event))
+    '''
+    with open(event) as f:
+        data = json.load(f)
+    print(data)
+    '''
+    #data = json.loads(json.dumps(event))
+    try:
+        data = data['result']
+    except:
+        data = data
+    df = pd.DataFrame(data['transactions'])
+
+    predictions = pr.main_predict(df, bucket_name, path)
+
+
+    #results = {'results': predictions} 
+    results_dicts = [json.loads(item) for item in predictions]
+    response = {
+        "statusCode": 200,
+        "headers": {
+            "Content-Type": "application/json"
+        },
+        "body": results_dicts
+    }
+    return response

09_model_pipeline/predict/src/predict.py

@@ -0,0 +1,14 @@
+from utils import * 
+import numpy
+class Predict:
+    def main_predict(data, bucket_name, path):
+        model = load_models_from_s3(bucket_name,  path )
+
+        probas = model.predict_proba(data)
+        preds = np.where(probas[:,1] >= 0.4, 1, 0)
+        data['prob_0'] = probas[:,0]
+        data['prob_1'] = probas[:,1]
+        data['predictions'] = preds
+        record = data.to_dict(orient='records')[0]
+        json_result = data.to_json(orient='records', lines=True).splitlines()
+        return json_result

09_model_pipeline/predict/src/utils.py

@@ -0,0 +1,15 @@
+import boto3
+from io import BytesIO
+import joblib
+def load_models_from_s3(bucket_name, object_name):
+
+    s3 = boto3.client('s3')
+    model_file = BytesIO()
+    s3.download_fileobj(bucket_name, object_name, model_file)
+
+    model_file.seek(0)
+
+
+    modelo = joblib.load(model_file)
+
+    return modelo