The Flow
Start by looking at the Flow. It illustrates in two different ways how to build sales forecasting models with historical weekly sales data for ten stores.
EXPLORE!In this reusable example project, we leverage Dataiku visual capabilities to build time series forecasting models using two different methods:
- The first one is based on statistical and Deep Learning time series models;
- The second one uses a more traditional Machine Learning approach.
Explore This Sample Project
Data Exploration
The project includes time series statistical analysis to better understand the structural pattern of the time series.
EXPLORE!
Build Statistical and Deep Learning model
In the first modeling approach, we leverage the sequential nature of our dataset and use a Visual Time Series Analysis to train a variety of statistical and Deep Learning models.
EXPLORE!
Use a Traditional Machine Learning Model
The second modeling approach uses visual recipes to enrich the dataset with date components, lagged features and store characteristics. This enriched dataset can then be used to train and deploy a Machine Learning Model.
EXPLORE!
Visualize your results with a dashboard
To communicate on the results, the project includes a dashboard with insights from the exploratory analysis, modeling phase, and updated forecast.
EXPLORE!