If you have an established data science practice, how do you get more from your AI investment? One answer is to augment your manually developed features with features discovered and evaluated automatically by dotData. Click above to watch our complimentary 30-minute video.
Video Transcript: Automated Feature Engineering for Enterprise ML
Good morning everyone. This is Ryohei, CEO at dotData. Today I’m going to talk about automated feature engineering for enterprise machine learning. Ai and machine learning are critical for financial services.
But here’s the thing, great machine learning does not promise great machine learning models. Machine learning models can be only as good as input data which is called the feature table. As you can see, you have a lot of data and you have to find important patterns as an input of machine learning. The secret of a successful machine learning project is basically about feature engineering. However, this feature engineering is difficult, time-consuming, and requires cross-domain knowledge.
It starts from hypothesizing features based on your domain expertise. Then data engineering comes in to transform your complex data into a single flat table. Of course, you cannot know which features hypothesis makes more sense to build machine learning models in advance. So that’s why data scientists must validate the statistical significance of each feature. This feature engineering has been 100% relying on intuition and experience before dotData.
Our key innovation is to automate this feature engineering. Transaction data, temporal data, geolocation data, tax data, and so on and so forth. You should have a lot of different types of data. Our AI engine automatically hypothesizes transforms and validates features and prepares AI features for you. You combine your domain features and our AI features to expand your feature space and build greater machine learning models. So what are the key values of our automated feature engineering?
First, it enhances your model accuracy. Again, machine learning models can be only as good as input features. Ai features augment your feature space and you can refine your machine learning models. Second, the module features engineering is good at finding core features based on your deep domain knowledge. On the other hand, automated feature engineering complement your ability to scan much more data set and explore much broader feature hypothesis.
That’s why you often find unknown and interesting patterns that you have never experimented with. Meanwhile, feature transparency is critical for enterprise machine learning. All AI features produced by dotData are designed to be transparent and interpretable. Our automated feature engineering is provided in two forms of product. One is dotData Py automated feature engine integrated with Python workflow.
If you have a Python data science team, dotData Py directly augmented enhance your machine learning workflow. Another product is dotData Enterprise. This is GUI based end-to-end automated machine learning platform combining auto feature engineering and automated machine learning. If you want to leverage your bi team to do a machine learning project, data enterprise provides no coding automation experience from Data through feature engineering to machine learning let me quickly introduce our customer success payment on billing customers. We covered 35% of decline transactions based on a smart dining model developed using our features.
This in fact corresponds to a 1% revenue increase. Another great success is a global property and casualty insurance. They achieved 250% contract rates of add-on clause using AI policy recommendation. Behind this AI policy recommendation system, hundreds of machine learning models enhanced by our features are running every day. This is the end of my presentation we are offering free proof of concept on trial.
So let us prove our automated feature engineering with your data science team. Thank you very much for your attention.
