Demand Forecasting with Deep Learning Models
In this notebook, we’ll demonstrate how to use Azure Machine Learning’s automated machine learning (AutoML) for time-series forecasting, specifically using deep learning models to forecast GitHub daily active users (DAU).
Prepare Data
We’ll start by loading the GitHub daily active users data into an MLTable, which is the required format for AutoML forecasting jobs.
import pandas as pd
from azure.ai.ml import Input
from azure.ai.ml.constants import AssetTypes
# Load the GitHub DAU data
df = pd.read_csv('github_dau.csv', index_col='date', parse_dates=['date'])
# Create an MLTable from the data
my_training_data_input = Input(
type=AssetTypes.MLTABLE,
path="./github_dau"
)
Configure AutoML Forecasting Job
Next, we’ll configure the AutoML forecasting job. We’ll enable the use of deep learning models by setting the enable_dnn_training
flag, and customize other parameters like the forecast horizon and time series ID columns.
from azure.ai.ml import automl
forecasting_job = automl.forecasting(
compute="gpu-cluster