Daytime Project

TAZI AutoML Platform

TAZI’s Automated Machine Learning is understandable continuous machine learning from data and humans, enables business domain experts to use machine learning to make predictions and take actions. It also helps data analysts and scientists for their daily model creation and deployment.

Data may change over time, TAZI AutoML uses multiple machine learning models, with supervised, unsupervised, semi-supervised algorithms, learning together and combined dynamically. While the batch ML performance degrades in time as data change, TAZI Continuous ML is able to get better in time. Compared to batch ML, robust continuous machine learning results in better accuracy and business benefits.


The business domain expert can inspect the explanation of the models in detail and suggest appropriate actions for each microsegment (pattern). For each instance an explanation is produced, so that the downstream BI systems can be configured to take the business actions automatically and domain experts are involved more in the creation and deployment of ML systems, making the system more robust, accurate and in line with business goals.


TAZI AutoML allows the business domain experts to provide instance or model level feedback to teach the machine learning system. This results in models learning faster while data changes. Including the domain expert feedback, the machine learning system needed labels only for the 23% of the data. This means less expert time labeling and less time waiting for the results to accumulate.


TAZI Data Profiler allows data analysts and business experts to examine and learn the statistics from their data. It tells which features should be imputed, ignored. In other words, they edit the data to get explainable models. Profiler makes recommendations to clean noisy data. Allows users to engineer new features. Auto feature selection is also provided so that the machine learning models are as simple and explainable as possible.


Hyperparameter Optimisation allows users to take multiple runs to find the best parameters for the model. Users can give multiple parameters to the models which would allow Tazi to find the best model as possible according to the business KPI needs.


TAZI can learn from streaming data, consisting of different types of inputs, such as numeric, discrete or text. Learning from streaming data reduces data transmission and storage costs. Inclusion of different types of data improves performance.




Contact Information: