implement forecasting start date poc

Author: jsn-li

evadb/README.md

@@ -1,6 +1,6 @@
 # Architecture of EvaDB Database System
 
-<img src="https://raw.githubusercontent.com/georgia-tech-db/evadb/master/docs/images/evadb/eva-arch.png" alt="EvaDB Architecture Diagram" width="500">
+<img src="https://raw.githubusercontent.com/georgia-tech-db/evadb/master/docs/images/evadb/evadb-arch.png" alt="EvaDB Architecture Diagram" width="500">
 
 * `server` - Code for launching server that sends client commands to command handler.
 * `parser` - Converts SQL queries to statements (e.g., CREATE, SELECT, INSERT, and LOAD statements).

evadb/executor/create_function_executor.py

@@ -19,6 +19,7 @@
 import pickle
 import re
 import time
+from datetime import datetime
 from pathlib import Path
 from typing import Dict, List
 
@@ -233,6 +234,8 @@ def handle_sklearn_function(self):
 
     def convert_to_numeric(self, x):
         x = re.sub("[^0-9.,]", "", str(x))
+        if x is None or x == '':
+            return x
         locale.setlocale(locale.LC_ALL, "")
         x = float(locale.atof(x))
         if x.is_integer():
@@ -422,6 +425,8 @@ def handle_forecasting_function(self):
         )  # shortens longer frequencies like Q-DEC
         season_length = season_dict[new_freq] if new_freq in season_dict else 1
 
+        start = arg_map.get("start", None)
+
         """
             Neuralforecast implementation
         """
@@ -683,6 +688,8 @@ def get_optuna_config(trial):
             FunctionMetadataCatalogEntry("horizon", horizon),
             FunctionMetadataCatalogEntry("library", library),
             FunctionMetadataCatalogEntry("conf", conf),
+            FunctionMetadataCatalogEntry("start", start),
+            FunctionMetadataCatalogEntry("frequency", new_freq),
         ]
 
         return (

evadb/functions/forecast.py

@@ -16,6 +16,7 @@
 
 import os
 import pickle
+import numpy as np
 
 import pandas as pd
 
@@ -39,6 +40,8 @@ def setup(
         horizon: int,
         library: str,
         conf: int,
+        start: str,
+        frequency: str,
     ):
         self.library = library
         if "neuralforecast" in self.library:
@@ -62,6 +65,8 @@ def setup(
         self.conf = conf
         self.hypers = None
         self.rmse = None
+        self.start = start
+        self.frequency = frequency
         if os.path.isfile(model_path + "_rmse"):
             with open(model_path + "_rmse", "r") as f:
                 self.rmse = float(f.readline())
@@ -70,12 +75,25 @@ def setup(
 
     def forward(self, data) -> pd.DataFrame:
         log_str = ""
+
+        dates = None
+        if self.start:
+            # dates = pd.DataFrame(columns=['unique_id', 'ds'])
+            # dates = dates.assign(unique_id=1)
+            dates = pd.date_range(start=self.start, periods=self.horizon, freq=self.frequency).to_numpy()
+
         if self.library == "statsforecast":
-            forecast_df = self.model.predict(
-                h=self.horizon, level=[self.conf]
-            ).reset_index()
+            if dates is None:
+                forecast_df = self.model.predict(
+                    h=self.horizon, level=[self.conf]
+                ).reset_index()
+            else:
+                forecast_df = self.model.fitted_[0, 0].forward(y=dates, h=self.horizon, level=[self.conf])
         else:
-            forecast_df = self.model.predict().reset_index()
+            if dates is None:
+                forecast_df = self.model.predict().reset_index()
+            else:
+                forecast_df = self.model.predict(futr_df=dates).reset_index()
 
         # Feedback
         if len(data) == 0 or list(list(data.iloc[0]))[0] is True: