Automated Machine Learning (AutoML) has revolutionized the world of Artificial Intelligence (IA) since its emergence in 1990. It’s generally known as a new approach of reducing data scientists' time-consuming tasks and a concept of automatizing AI processes.
It is on this innovative approach that SmartPredict has ingeniously enriched its feature for its first version.
All this to enable its users — whether they are citizens (novice) or expert data scientists — to be more efficient in the realization of their AI project.
They will benefit from an AI platform that will perform for them — in a matter of a few minutes — the tedious tasks of their AI projects that professionals would struggle to complete over several days. But above all, the flexibility aspect has been highlighted in its feature so that they have full control of their AI project.
To better enlighten you on what it’s all about, this article describes the main features of Autoflow (the SmartPredict’s feature based on AutoML 2.0).
1) Easy to use for all users
Users, whatever their expertise domain (data analytics, management, sales, marketing, etc), regardless of their AI knowledge, are able to successfully realize end-to-end production-ready AI projects, thanks to Autoflow.
They have at their disposal an intuitive interface that guides them to perform their AI project until they get a prediction or forecast, or the expected result from ML/DL models. There is no question of coding, nor technical aspect to handle but just straightforward configuration. The steps involved are: importing data, defining the target value to predict, waiting for Autoflow to perform all processes, and getting predictions directly in the same platform. For more purposes, the prediction can be implemented in other business software by using provided credentials.
The interface to configure the project with Autoflow
2) Automated performant engine
Technically speaking, the automation process behind Autoflow is fast, powerful, and transparent. As mentioned before, users have just to define the problem to solve and Autoflow smartly performs in their place all the AI processes automatically until providing reliable results.
The processes involved are :
- Auto data Preprocessing: a data mining technique that transforms data into an understandable format. In this step, Autoflow inspects the user's data and performs the needed process to have complete, consistent, and clean data.
- Auto Feature Engineering: a process that generates new features from data to increase the model performance. Autoflow does not only produce computable data, but it also acquires deep insight to generate relevant features that will contribute to a more performant AI model.
- Auto Model Selection: according to the user-defined problem configuration, Autoflow automatically detects the ML or DL models that match the problem requirements. In this step, Autoflow trains and fine-tuns many models and presents the best models classified by their metrics in the end. Users just have to select the best one.
- Build the resulted ML workflow: once selected, Autoflow generates the resulted ML workflow — in the form of a customizable flowchart — which will train the selected ML / DL models
The automation processes performed by Autoflow
The resulting Autoflow's work
3) Auto Generation of 100% customizable ML workflow
AI experts can have full control of their AI projects. Autoflow generates the results of its automation work in the form of a customizable flowchart. Thus, they are free to bring their expertise, whether to modify the modules' configuration, or create their own model with Python code, or test other existing models in the platform. In any case, it saves them a lot of time and makes them more efficient, as they can benefit from some automated processes that Autoflow has performed.
The generated customizable flowchart (which trains and evaluates the selected ML/DL model)
The generated deployment flowchart (which is also customizable)
4) Prediction ready
It is well worth that Autoflow has performed users' time-consuming tasks for their AI projects. But that's not all, they can directly test the deployed model, and get the prediction from it directly, in the SmartPredict Predict space.
At the time of speaking, SmartPredict Autoflow mainly manages end-to-end time series forecasting AI projects: you just have to insert your historical dataset and get a forecasting value.
It works for sales forecasting, weather forecasting, stock price forecasting, and many more.
So what are you waiting for? Try it now and get the most of AI technology.