Have your datasets team up with predictive modeling to boost up your business in an easy way.
Your data must be stored in a Tabular way where each column is a variable (sometimes also called features) and each row is a data point (or data samples). The first line must contain the name of the variables.
Once you have created a project and imported your data, you can check everything is fine. TADA automatically detects if there is any missing value and also detects automatically the type of the variables would be a number, a category or simply a text.
It helps you to start analyzing your data immediately.
You need then to define a “Goal” which is the variable you want to understand and possibly even predict in the future. There are 3 types of Goals: Binary (yes/no questions), Multi-Class (between 3 and 50 categories) & Regression (for numbers).
Note that a “Project” can have several different “Goals”
Your dataset may contain variables that are useless to understand your Goal. TADA’s automatic variable reduction removes them and even shows you what are the most important variables for your “Goal”.
Also you may want to play with the tool and exclude one variable simply to see how the system would work without this information. You can do that manually.
Note that a “Goal” can have several different “Variable Set”
Once your Goal is defined and you have selected the variables you want to use, it is time to let TADA analyze your data. TADA does that by generating thousands of different “predictive models” and it selects the best of them. A predictive model “summarizes” the data: it tries to explain how your “Goal” can be deduced from all the other variables.
You can play with the parameters to have the first rough results in a few seconds and then create another model that will take several minutes to favor accuracy over speed.
Note: TADA models are well known to be very accurate and at the same time simple enough to be understood by humans.
Once the model is generated, we provide loads of metrics that measure how good it is: accurate, precise, etc…. These metrics are well known among data scientists but if you need to be refreshed of what they measure you have all the explanations directly in the UI.
Do not hesitate to share them with your data science team!
One of the most interesting features of TADA is called Sensitivity Analysis. It explains how a variable influences the final result. It helps you to understand with accuracy how your “Goal” behaves. Many of our customers have understood some key phenomenons based on this simple feature.
If you have generated a good – read accurate – model, you can then play with it. You can predict your goal based on all the other key variables. Most find it difficult to stop playing with it!
*will be available soon
Start your AI path to performance
Test easily TADA with our test data here:
ZGP : AI and mathematics at the service of your DATA.
ZGP combines two main fields of today’s AI: Symbolic Regression and Evolutionary Programming, to reach Zoetrope Genetic Programming achievement. It creates simple mathematical expressions that are particularly good at predicting or classifying Small Data. When most of today’s solutions take hours and hours (and a large amount of data) to produce decent result, ZGP produces much better models at a much faster pace.
After 10 years of research in AI, we continue to innovate.
We have now partnered with major research institutes (INRIA) to accelerate our research. We continue to invest massively in research and have built partnership with some of the most renowned mathematicians and researchers in the field. Like our algorithms, we evolve!
Multiple modeling & discriminating capacities
Binary classification, Regression and Multi-class classification modeling are available. The modeling algorithm is able to consider a large number of variables during analysis and automatically select a minimal subset comprised of the most useful variables.
Minimalistic & efficient
Models produced are minimalistic in the sense of having a minimum reliance upon larger quantities of independent variables. Models have a peak maximum efficiency when employing 3 to 7 independent variables. The algorithm is able to discern physical signals in small amounts of data, i.e very few rows
Understandable & insightful
Models produced are in the form of a human readable mathematical equation which can be deployed in computing languages (Java, C++). Having specific equations describing system behaviors enables both wide application and detailed exploration of underlying phenomena.
The algorithm produces accurate estimation of future model performance. The accuracy realized when a model is deployed closely matches that estimated by the modeling process