Added api file

Dockerfile

@@ -0,0 +1,17 @@
+# 
+FROM python:3.9
+
+# 
+WORKDIR /code
+
+# 
+COPY ./requirements.txt /code/requirements.txt
+
+# 
+RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
+
+# 
+COPY ./src /code/src
+
+# 
+CMD ["uvicorn", "src.app.app:app", "--host", "0.0.0.0", "--port", "7860"]

requirements.txt

@@ -0,0 +1,10 @@
+matplotlib==3.3.4
+numpy==1.22.4
+pandas==1.2.4
+scipy==1.6.2
+seaborn==0.11.1
+scikit-learn==0.24.1
+xgboost==1.7.3
+streamlit==1.23.1
+fastapi[all]==0.98.0
+uvicorn[standard]==0.22.0

src/app/app.py

@@ -0,0 +1,77 @@
+from fastapi import FastAPI
+import uvicorn
+from datetime import datetime
+from typing import Annotated
+import os
+import sys
+import datetime
+import pandas as pd
+
+sys.path.append(os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))))
+from src.utils import load_file, make_predcition, date_extracts
+
+
+
+
+
+# Create an instance of FastAPI
+app = FastAPI(debug=True)
+
+# get absolute path
+DIRPATH = os.path.dirname(os.path.realpath(__file__))
+
+# set path for ml files
+ml_contents_path = os.path.join(DIRPATH, '..', 'assets', 'ml_components', 'toolkit_folder')
+
+# get contents
+ml_contents = load_file(ml_contents_path)
+
+Encoder = ml_contents["OneHotEncoder"]
+model = ml_contents["model"]
+features_ = ml_contents['feature_names']
+
+
+
+# define endpoints
+
+@app.get('/')
+def root():
+    return 'Welcome to the Gorecery Sales Forecasting API'
+
+@app.get('/health')
+def check_health():
+    return {'status': 'ok'}
+
+@app.post('/predict')
+async def predict_sales( store_id: int, category_id: int, onpromotion: int,
+                  city: str, store_type: int, cluster: int, date_: Annotated[datetime.date, "The date of sales"] = datetime.date.today()):
+
+    # create a dictionary of inputs
+    input = {
+    'store_id':[store_id], 
+    'category_id':[category_id], 
+    'onpromotion' :[onpromotion],
+    'type' : [store_type], 
+    'cluster': [cluster],
+    'city' : [city],
+    'date_': [date_]
+    }   
+
+    # convert to dataframe and extract datetime features
+    input_data = pd.DataFrame(input)
+    date_extracts(input_data)
+
+
+    # make prediction
+    sales = make_predcition(Encoder, model, input)
+    sales_value = float(sales[0])
+    return {'sales': sales_value}
+
+
+
+if __name__ == "__main__":
+    uvicorn.run('app:app', reload=True)
+
+
+
+

src/utils.py

@@ -0,0 +1,71 @@
+import pandas as pd
+import numpy as np
+import pickle
+from io import StringIO
+from functools import lru_cache
+
+
+def return_features():
+    features_ = ['store_id', 'category_id', 'onpromotion', 'type', 'cluster', 'year', 'month', 'dayofmonth', 'dayofweek', 'dayofyear', 'weekofyear', 'quarter', 'is_month_start', 'is_month_end', 'is_quarter_start', 'is_quarter_end', 'is_year_start', 'is_year_end', 'year_weekofyear', 'x0_Accra', 'x0_Aflao', 'x0_Akim Oda', 'x0_Akwatia', 'x0_Bekwai', 'x0_Cape coast', 'x0_Elmina,', 'x0_Gbawe', 'x0_Ho', 'x0_Hohoe', 'x0_Kintampo', 'x0_Koforidua', 'x0_Kumasi', 'x0_Mampong', 'x0_Obuasi', 'x0_Prestea', 'x0_Suhum', 'x0_Tamale', 'x0_Techiman', 'x0_Tema', 'x0_Teshie', 'x0_Winneba']
+    return features_
+
+
+
+@lru_cache(maxsize=100, )
+def load_file(filename):
+    with open(filename, 'rb') as file: # read file
+        contents = pickle.load(file) # load contents of file
+    return contents
+
+def date_extracts(df):
+    # Extract date features
+    df['date_'] = pd.to_datetime(df['date_'], errors='coerce')
+    df['year'] = df['date_'].dt.year
+    df['month'] = df['date_'].dt.month
+    df['dayofmonth'] = df['date_'].dt.day
+    df['dayofweek'] = df['date_'].dt.dayofweek
+    df['dayofyear'] = df['date_'].dt.dayofyear
+    df['weekofyear'] = df['date_'].dt.weekofyear
+    df['quarter'] = df['date_'].dt.quarter
+    df['is_month_start'] = df['date_'].dt.is_month_start.astype(int)
+    df['is_month_end'] = df['date_'].dt.is_month_end.astype(int)
+    df['is_quarter_start'] = df['date_'].dt.is_quarter_start.astype(int)
+    df['is_quarter_end'] = df['date_'].dt.is_quarter_end.astype(int)
+    df['is_year_start'] = df['date_'].dt.is_year_start.astype(int)
+    df['is_year_end'] = df['date_'].dt.is_year_end.astype(int)
+    df['year_weekofyear'] = ((df['year'] -2017) *100 )+ df['weekofyear']
+
+    df.drop(columns=['date_'], inplace=True)
+
+
+
+
+# defining categories and numeric columns
+def make_predcition(Encoder, model, input_df):
+    if isinstance(input_df, dict):
+        # Put the input dictionary in a dataset
+        input_data = pd.DataFrame(input_df)
+    
+    if isinstance(input_df, pd.DataFrame):
+        input_data = input_df
+    col = ['city']
+    columns = list(input_data.columns)
+    encoded_cat = Encoder.transform(input_data[col])
+    encoded_cols = Encoder.get_feature_names()
+    encoded_cat_ = pd.DataFrame(encoded_cat, columns=encoded_cols)
+
+
+
+    
+    # we dropped the categorical encoder column before we concat 
+    train_enc = input_data.drop(['city'],axis = 1)
+    input_d = pd.concat([train_enc, encoded_cat_], axis=1)
+    input_d = input_d.reindex(columns=return_features())
+
+    # print(input_d)
+
+    # # convert input_data to a numpy array before flattening to convert it back to a 2D array
+    # input_df= input_d.to_numpy()
+    print(model.get_booster().feature_names)
+    prediction = model.predict(input_d)
+    return prediction