Merge pull request #107 from Hestia-Homes/carbon-dev-model

Carbon dev model

MODEL_REGISTRY.md

@@ -16,17 +16,17 @@
         "active": true
     },
     "heat": {
-        "version": "v0.3.0",
+        "version": "v0.4.0",
         "stage": {
-            "dev": "v0.3.0"
+            "dev": "v0.4.0"
         },
         "registered": true,
         "active": true
     },
     "carbon": {
-        "version": "v0.3.0",
+        "version": "v0.4.0",
         "stage": {
-            "dev": "v0.3.0"
+            "dev": "v0.4.0"
         },
         "registered": true,
         "active": true

deployment/Dockerfile.prediction.lambda

@@ -9,7 +9,7 @@ ARG RUNTIME_ENVIRONMENT
 ENV RUNTIME_ENVIRONMENT=${RUNTIME_ENVIRONMENT}
 
 # Install necessary build tools - required to test locally
-RUN yum install -y gcc python3-devel
+RUN yum install -y gcc python3-devel gcc-c++
 
 # Install python packages
 COPY modules/ml-pipeline/src/pipeline/requirements/predictions/requirements.txt ./requirements.txt

modules/ml-pipeline/src/README.md

@@ -1,3 +1,3 @@
-# The generic reproducible ML-pipeline
+# The generic reproducible ML-pipeline!
 
 Pipeline required to build a model to produce an output, that gets hashed via DVC

modules/ml-pipeline/src/pipeline/.gitignore

@@ -1,3 +1,4 @@
 
 # Ignore dynaconf secret files
 .secrets.*
+example.py

modules/ml-pipeline/src/pipeline/configs/build_model.yaml

@@ -19,3 +19,4 @@ default:
       excluded_model_types: ['RF', 'FASTAI', 'CAT', 'NN_TORCH', 'KNN', 'XT']
       infer_limit: 0.05
       infer_limit_batch_size: 10000
+      ag_args_ensemble: {'num_folds_parallel': 2}

modules/ml-pipeline/src/pipeline/configs/feature_processor_logic.py

@@ -18,30 +18,44 @@ def remove_starting_columns(df):
     return df
 
 
-def remove_floor_height_ending(df):
-    # df.describe(percentiles=[0.005,0.99])['FLOOR_HEIGHT_ENDING']
-    # shows bottom 0.5 percentile is 1.665
-    # So keep anything above this
-    df = df[df["floor_height_ending"] > 1.665].reset_index(drop=True)
-    print("we in here")
+def keep_negative_heat_change(df):
+    df = df[df["heat_demand_change"] < 0]
     return df
 
 
-def remove_minimum_habitable_room_size(df):
-    # Need minimum of 6.5m per habitable room
-    df = df[
-        df["total_floor_area_ending"] / df["number_habitable_rooms"] > 6.5
-    ].reset_index(drop=True)
+def keep_non_negative_carbon_ending(df):
+    df = df[df["carbon_ending"] > 0]
     return df
 
 
-def keep_flats(df):
-    df = df[df["property_type"] == "Flat"]
+def keep_negative_carbon_change(df):
+    df = df[df["carbon_change"] < 0]
     return df
 
 
-def keep_non_zero_rdsap(df):
-    df = df[df["rdsap_change"] != 0]
+# TODO: Move to ETL pipeline
+def remove_unreasonable_habitable_rooms(df):
+    """
+    Assumption is that proportion of floor area to habitable rooms should be at least 6.5m2
+    """
+    minimum_room_size_index = (
+        df["total_floor_area_ending"] / df["number_habitable_rooms"] >= 6.5
+    )
+    df = df[minimum_room_size_index]
+    return df
+
+
+def remove_top_1_percent_heat_demand(df):
+    # threshold_value = df.describe(percentiles=[0.99])['HEAT_DEMAND_STARTING']['99%']
+    threshold_value = 860
+    df = df[df["heat_demand_starting"] < threshold_value]
+    return df
+
+
+def remove_top_1_percent_carbon(df):
+    # threshold_value = df.describe(percentiles=[0.99])['CARBON_STARTING']['99%']
+    threshold_value = 18
+    df = df[df["carbon_starting"] < threshold_value]
     return df
 
 
@@ -54,10 +68,12 @@ def keep_non_zero_rdsap(df):
 #     return df
 
 business_logic = {
-    # "keep_non_zero_rdsap": keep_non_zero_rdsap,
-    # "keep_flats": keep_flats,
-    # "remove_minimum_habitable_room_size": remove_minimum_habitable_room_size,
-    # "remove_floor_height_ending": remove_floor_height_ending
+    "remove_unreasonable_habitable_rooms": remove_unreasonable_habitable_rooms,
+    "keep_negative_heat_change": keep_negative_heat_change,
+    "keep_negative_carbon_change": keep_negative_carbon_change,
+    "remove_top_1_percent_heat_demand": remove_top_1_percent_heat_demand,
+    "remove_top_1_percent_carbon": remove_top_1_percent_carbon,
+    "keep_non_negative_carbon_ending": keep_non_negative_carbon_ending,
     # "remove_starting_columns": remove_starting_columns
     # "keep_ENDING_COLUMNS": keep_ending_columns
 }

modules/ml-pipeline/src/pipeline/configs/post_prediction_logic.py

@@ -1,23 +1,24 @@
 """
 After predictions, we may want to apply some post processing to the predictions
 """
+
 import pandas as pd
 
 
 def clip_predictions_to_minimum_value(
-    data: pd.DataFrame, predictions: pd.Series, minimum_value: int = 0
+    data: pd.DataFrame,
+    predictions: pd.Series,
 ) -> pd.Series:
 
     series_name = predictions.name
     predictions.name = "predictions"
+    predictions = predictions.astype(data["carbon_starting"].dtype)
     predictions_df = pd.concat([data, predictions], axis=1)
     # We expect all prediction to be atleast one point improvement
-    replace_index = (
-        predictions_df["sap_starting"] + minimum_value > predictions_df["predictions"]
-    )
-    predictions_df.loc[replace_index, "predictions"] = (
-        predictions_df.loc[replace_index, "sap_starting"] + minimum_value
-    )
+    replace_index = predictions_df["predictions"] > predictions_df["carbon_starting"]
+    predictions_df.loc[replace_index, "predictions"] = predictions_df.loc[
+        replace_index, "carbon_starting"
+    ]
 
     predictions_new = predictions_df["predictions"]
     predictions_new.name = series_name

modules/ml-pipeline/src/pipeline/configs/settings.yaml

@@ -18,13 +18,8 @@ default:
   prepare_data:
     input_dataclient_type: aws-s3
     output_dataclient_type: local
-    # data_filepath: s3://retrofit-data-dev/sap_change_model/dataset_with_differencing.parquet
-    # data_filepath: s3://retrofit-data-dev/sap_change_model/floor_area_clean_test.parquet
-    # data_filepath: s3://retrofit-data-dev/sap_change_model/dataset_without_differencing.parquet
-    # data_filepath: s3://retrofit-data-dev/sap_change_model/dataset_test.parquet
-    # data_filepath: s3://retrofit-data-dev/sap_change_model/dataset.parquet
-    data_filepath: s3://retrofit-datalake-dev/dataset_with0perm_all.parquet
-    train_proportion: 1
+    data_filepath: s3://retrofit-data-dev/sap_change_model/2024-03-22-18-56-53/dataset_rooms.parquet
+    train_proportion: 0.9
     output_train_filepath: ./data/prepared_data/train.parquet
     output_test_filepath: ./data/prepared_data/test.parquet
 
@@ -33,9 +28,13 @@ default:
     feature_processor_config:
       subsample_amount: null
       subsample_seed: 0
-      target: sap_ending
+      target: carbon_ending
       identifier_columns: ["uprn"]
-      drop_columns: ["heat_demand_change", "carbon_change", "rdsap_change", "heat_demand_ending", "carbon_ending"]
+      # drop_columns: ["heat_demand_change", "carbon_change", "rdsap_change", "heat_demand_ending", "sap_ending"]
+      drop_columns: [
+        "heat_demand_change", "carbon_change", "rdsap_change", "heat_demand_ending", "sap_ending", "days_to_starting", "days_to_ending",
+        'number_habitable_rooms_starting', 'number_habitable_rooms_ending', 'number_heated_rooms_starting', 'number_heated_rooms_ending',
+        'number_habitable_rooms', 'number_heated_rooms']
       # retain_features: ["SAP_STARTING", "TOTAL_FLOOR_AREA_DIFF"]
       retain_features: null
 

modules/ml-pipeline/src/pipeline/core/MLModels.py

@@ -25,7 +25,7 @@ def model_factory(model_type: str) -> MLModel:
     models = {
         "SKLearnLinearRegression": SKLearnLinearRegression(),
         "SKLearnSVMRegression": SKLearnSVMRegression(),
-        "AutogluonAutoML": AutogluonAutoML()
+        "AutogluonAutoML": AutogluonAutoML(),
         # ADD OTHER MODELS HERE
     }
 
@@ -151,6 +151,7 @@ class AutogluonAutoML:
         "excluded_model_types",
         "infer_limit",
         "infer_limit_batch_size",
+        "ag_args_ensemble",
     ]
 
     def load_model(self, path: Union[Path, str]) -> None:
@@ -207,6 +208,7 @@ class AutogluonAutoML:
             excluded_model_types=model_hyperparameters["excluded_model_types"],
             infer_limit=model_hyperparameters["infer_limit"],
             infer_limit_batch_size=model_hyperparameters["infer_limit_batch_size"],
+            ag_args_ensemble=model_hyperparameters["ag_args_ensemble"],
         )
 
     def predict(

modules/ml-pipeline/src/pipeline/dvc.lock

@@ -1,12 +1,23 @@
 schema: '2.0'
 stages:
+  startup_cleanup:
+    cmd: python 0_startup_cleanup.py
+    deps:
+    - path: 0_startup_cleanup.py
+      hash: md5
+      md5: b1b12f6b6393fbf8b83d23684df0a3d4
+      size: 1220
+    params:
+      configs/settings.yaml:
+        default.startup_cleanup.artefacts: ./data
+        default.startup_cleanup.metrics: ./metrics
   prepare_data:
     cmd: python 1_prepare_data.py
     deps:
     - path: 1_prepare_data.py
       hash: md5
-      md5: 1793a35e71751d3c84f9affc67ecb9a8
-      size: 4296
+      md5: 11a3b8bfdfe199ab7ecc39ccc5652649
+      size: 4298
     params:
       configs/settings.yaml:
         default.feature_processor.feature_processor_config.drop_columns:
@@ -14,23 +25,32 @@ stages:
         - carbon_change
         - rdsap_change
         - heat_demand_ending
-        - carbon_ending
+        - sap_ending
+        - days_to_starting
+        - days_to_ending
+        - number_habitable_rooms_starting
+        - number_habitable_rooms_ending
+        - number_heated_rooms_starting
+        - number_heated_rooms_ending
+        - number_habitable_rooms
+        - number_heated_rooms
         default.feature_processor.feature_processor_config.retain_features:
         default.feature_processor.feature_processor_config.subsample_amount:
         default.feature_processor.feature_processor_config.subsample_seed: 0
-        default.feature_processor.feature_processor_config.target: sap_ending
+        default.feature_processor.feature_processor_config.target: carbon_ending
         default.feature_processor.feature_processor_type: dataframe
-        default.prepare_data.data_filepath: s3://retrofit-datalake-dev/dataset_with0perm_all.parquet
+        default.prepare_data.data_filepath:
+          s3://retrofit-data-dev/sap_change_model/2024-03-22-18-56-53/dataset_rooms.parquet
         default.prepare_data.input_dataclient_type: aws-s3
         default.prepare_data.output_dataclient_type: local
         default.prepare_data.output_test_filepath: ./data/prepared_data/test.parquet
         default.prepare_data.output_train_filepath: ./data/prepared_data/train.parquet
-        default.prepare_data.train_proportion: 1
+        default.prepare_data.train_proportion: 0.9
     outs:
     - path: data/prepared_data/
       hash: md5
-      md5: 84fa631bd02686b052d6a7144eafd38e.dir
-      size: 43859225
+      md5: 824541f44e6538d2ef10e9d754c79743.dir
+      size: 36691842
       nfiles: 2
   build_model:
     cmd: python 2_build_model.py
@@ -41,8 +61,8 @@ stages:
       size: 4820
     - path: data/prepared_data
       hash: md5
-      md5: 84fa631bd02686b052d6a7144eafd38e.dir
-      size: 43859225
+      md5: 824541f44e6538d2ef10e9d754c79743.dir
+      size: 36691842
       nfiles: 2
     params:
       configs/build_model.yaml:
@@ -70,21 +90,23 @@ stages:
               - XT
               infer_limit: 0.05
               infer_limit_batch_size: 10000
+              ag_args_ensemble:
+                num_folds_parallel: 2
     outs:
     - path: data/fit_predictions/
       hash: md5
-      md5: ede187e9d0bffdef054f573f3c2bd222.dir
-      size: 3578590
+      md5: 5a3091120d3497fa00b994d91bc7e5eb.dir
+      size: 3664806
       nfiles: 1
     - path: data/model/
       hash: md5
-      md5: b2ad0b538dc4aef0de3d431fc9c40c4f.dir
-      size: 814720415
+      md5: 074da8dcfa515b9f3d082b21c7d76616.dir
+      size: 721558897
       nfiles: 31
     - path: metrics/fit_metrics.json
       hash: md5
-      md5: c45b84f12971a0156e4f3d85d3e725f5
-      size: 218
+      md5: 728a49dcef5a98182325df455f929a33
+      size: 225
   generate_predictions:
     cmd: python 3_generate_predictions.py
     deps:
@@ -94,13 +116,13 @@ stages:
       size: 2464
     - path: data/model
       hash: md5
-      md5: b2ad0b538dc4aef0de3d431fc9c40c4f.dir
-      size: 814720415
+      md5: 074da8dcfa515b9f3d082b21c7d76616.dir
+      size: 721558897
       nfiles: 31
     - path: data/prepared_data
       hash: md5
-      md5: 84fa631bd02686b052d6a7144eafd38e.dir
-      size: 43859225
+      md5: 824541f44e6538d2ef10e9d754c79743.dir
+      size: 36691842
       nfiles: 2
     params:
       configs/settings.yaml:
@@ -112,8 +134,8 @@ stages:
     outs:
     - path: data/predictions/
       hash: md5
-      md5: 5e60ca251af51de6fef3d0c659f8bb27.dir
-      size: 627416
+      md5: 680f51234d214d4cab9e6a064c75fc5d.dir
+      size: 499546
       nfiles: 1
   generate_metrics:
     cmd: python 4_generate_metrics.py
@@ -124,13 +146,13 @@ stages:
       size: 3484
     - path: data/predictions
       hash: md5
-      md5: 5e60ca251af51de6fef3d0c659f8bb27.dir
-      size: 627416
+      md5: 680f51234d214d4cab9e6a064c75fc5d.dir
+      size: 499546
       nfiles: 1
     - path: data/prepared_data
       hash: md5
-      md5: 84fa631bd02686b052d6a7144eafd38e.dir
-      size: 43859225
+      md5: 824541f44e6538d2ef10e9d754c79743.dir
+      size: 36691842
       nfiles: 2
     params:
       configs/settings.yaml:
@@ -140,16 +162,5 @@ stages:
     outs:
     - path: metrics/metrics.json
       hash: md5
-      md5: 033efa4d4044b6b6fc92dd37194727fa
-      size: 225
-  startup_cleanup:
-    cmd: python 0_startup_cleanup.py
-    deps:
-    - path: 0_startup_cleanup.py
-      hash: md5
-      md5: b1b12f6b6393fbf8b83d23684df0a3d4
-      size: 1220
-    params:
-      configs/settings.yaml:
-        default.startup_cleanup.artefacts: ./data
-        default.startup_cleanup.metrics: ./metrics
+      md5: 67b7ab30a4b0839d20bc6eb0c84e4dd1
+      size: 226

modules/ml-pipeline/src/pipeline/requirements/predictions/requirements-dev.txt

@@ -1,7 +1,7 @@
 joblib==1.3.2
 boto3==1.28.17
-pandas==1.5.3
-autogluon==0.8.2
-dynaconf==3.2.0
+pandas==2.1.4
+autogluon==1.0.0
+dynaconf==3.2.1
 pyarrow==13.0.0
 pre-commit==3.3.3

modules/ml-pipeline/src/pipeline/requirements/predictions/requirements.txt

@@ -1,7 +1,7 @@
 joblib==1.3.2
 boto3==1.28.17
-pandas==1.5.3
-autogluon==0.8.2
-dynaconf==3.2.0
+pandas==2.1.4
+autogluon==1.0.0
+dynaconf==3.2.1
 pyarrow==13.0.0
 PyYAML==6.0.1

modules/ml-pipeline/src/pipeline/requirements/training/requirements-dev.txt

@@ -1,9 +1,10 @@
 joblib==1.3.2
 boto3==1.28.17
-pandas==1.5.3
-autogluon==0.8.2
-dynaconf==3.2.0
-alibi==0.9.4
+pandas==2.1.4
+autogluon==1.0.0
+ray==2.6.3
+dynaconf==3.2.1
+alibi==0.9.5
 shap==0.42.1
 pyarrow==13.0.0
 pre-commit==3.3.3

modules/ml-pipeline/src/pipeline/requirements/training/requirements.txt

@@ -1,4 +1,4 @@
 boto3==1.28.41
-pandas==1.5.3
-autogluon==0.8.2
-dynaconf==3.2.0
+pandas==2.1.4
+autogluon==1.0.0
+dynaconf==3.2.1