The Evolution, Misconceptions, and Reality of AutoML

With every new technology, especially in the early days, comes a share of misconceptions, fallacy, and ambiguity. That’s why our CEO Ryohei Fujimaki shares the top five myths and reality of AutoML with RTInsights.

Businesses are looking for ways to make AI more accessible by automating complex manual data science processes. Data scientists are typically hired by Fortune 500 companies and work for a handful of companies. Most companies struggle to find and retain data scientists. Businesses can empower citizen data scientists to make data-driven decisions using advanced analytical tools that automate complex manual data-science processes. The focus of first-generation AutoML platforms has been on building and validating models automatically, but next-generation AutoML platforms can automate the entire data science process, including feature engineering and data preparation, to build models in days instead of months.

Feature engineering entails building features using relational, transactional, temporal, geo-locational, or text data to test hypotheses, develop and evaluate ML models, and repeat the process until the results become acceptable for businesses. Data is spread across multiple databases in multiple formats not suitable for analytics. It is also spread across various databases in various forms and requires data preparation tools to help with data cleansing, formatting, standardization, and additional table joining, and further data preparation. The traditional data science approach is to assumes model accuracy is more important than feature transparency and explanation. However, white-box models are preferred in many enterprise use cases because they clearly explain how the model works and how it produces predictions.

The AutoML 2.0 revolution in business analytics will enable anyone to embark on predictive analytics projects without requiring a data science background. The AutoML Platform makes it possible to deploy an ML model in production environments with minimum effort and maximum impact and enables enterprise data science and software/IT teams to operationalize complex data science projects. AutoML 2.0 platforms automate the entire data science process and can provide better insights faster than ever before. Data science can be made ‘citizen’ for each project and be interpreted more quickly by domain experts, providing the shortest time-to-market for new data-driven business applications. Read the full article at RTInsights