I am using this solution to deploy a POC for the forecasting solution. One of the demand from the customers is to run a daily forecast vs a weekly forecast.
I've made the below changes to the code in the file 'Sales_Data_Agrgregation_2.0_Blob.py file. I do not see the predictions being accurate. Could you please confirm if the changes are accurate?
########################## 2.2 aggregate the sales data and join with store and product attributes
create week_start for df_sales
df_sales_date_min=processed_time_d[0]
##I do not "roll" the daily dates to be identified as a week. I leave the dates as is.
df_sales=df_sales.withColumn('week_start',col('date_date'))
df_sales=df_sales.groupBy(['week_start','store_id','product_id']).agg({"*": "count"}).withColumnRenamed('count(1)', 'sales')
## join the df_sales with df_price_change, warning: under this scenario, since every week each product in each department in each store has an entry in df_price_change will this accurately reflect the daily prediction?
df_sales=df_sales.join(df_price_change,["week_start","store_id","product_id"],'left_outer')
## join df_sales with df_products and df_stores to get products and stores attributes
df_sales=df_sales.join(df_products,["product_id"],"inner").join(df_stores_join,["store_id"],"inner")
Hi
I am using this solution to deploy a POC for the forecasting solution. One of the demand from the customers is to run a daily forecast vs a weekly forecast.
I've made the below changes to the code in the file 'Sales_Data_Agrgregation_2.0_Blob.py file. I do not see the predictions being accurate. Could you please confirm if the changes are accurate?
########################## 2.2 aggregate the sales data and join with store and product attributes
create week_start for df_sales
df_sales_date_min=processed_time_d[0]
##I do not "roll" the daily dates to be identified as a week. I leave the dates as is. df_sales=df_sales.withColumn('week_start',col('date_date')) df_sales=df_sales.groupBy(['week_start','store_id','product_id']).agg({"*": "count"}).withColumnRenamed('count(1)', 'sales')
## join the df_sales with df_price_change, warning: under this scenario, since every week each product in each department in each store has an entry in df_price_change will this accurately reflect the daily prediction?
df_sales=df_sales.join(df_price_change,["week_start","store_id","product_id"],'left_outer') ## join df_sales with df_products and df_stores to get products and stores attributes df_sales=df_sales.join(df_products,["product_id"],"inner").join(df_stores_join,["store_id"],"inner")