Digital transformation is largely driven by data. And businesses today are looking for as many ways as possible to derive as much insight as they can from their data. Indeed, machine learning (ML) has become a rapidly increasing presence across industries in recent years. In 2017, ML’s impact on powering applications and services was huge for companies such as Microsoft, Google, and Amazon. And ML’s usefulness continues to grow in businesses of all sizes: examples include fraud prevention customer service chatbots at banks, automatically targeting segments of the market at marketing agencies, and e-commerce product recommendations and personalization by retailers.
While ML certainly is a hot topic, there is another trend that is becoming just as popular: automated machine learning (autoML).
According to TDWI, the field of autoML is evolving so quickly, there is no universally agreed-upon definition. Basically, autoML gives ML experts tools to automate repetitive tasks by applying ML to ML itself. According to Google Research, the goal of automating ML is to create techniques for computers to solve new ML problems automatically, without needing human ML experts to intercede on each new problem. This capability will result in truly intelligent systems.
Further, opportunities are being created thanks to autoML. After all, these types of technologies require expert researchers, data scientists and engineers, and across the world, yet these positions are in short supply. In fact, these positions are so poorly filled that the “citizen data scientist” has come into being. Rather than a direct replacement, this complementary role hires people who lack specific advanced data scientist expertise. Yet, they can generate models using state-of-the-art diagnostic and predictive analytics. This capability is due to the rise of autoML, which can automate many of the tasks once performed by data scientists.
To compensate for the lack AI/ML experts, autoML has the ability to automate some of the more repetitive tasks of ML while boosting the productivity of data scientists. Tasks that can be automated include choosing data sources, feature selection, and data prep, which frees up time for marketing and business analysts to focus on essential tasks. For instance, data scientists can fine-tune more new algorithms, build more models in less time, and improve model quality and accuracy.
According to the Harvard Business Review, organizations have shifted toward amplifying predictive power. To do so, they have paired big data with complex automated ML. AutoML is advertised as creating opportunities to democratize ML by allowing firms with limited data science expertise to develop analytical pipelines that have the ability to solve sophisticated business problems.
To illustrate how this works, a common ML pipeline is made up of pre-processing, feature extraction, feature selection, feature engineering, algorithm selection, and hyper-parameter tuning. But there’s a high barrier to entry due to the considerable expertise and time it takes to implement these steps.
One of the benefits of autoML is that it removes some of these constraints by significantly reducing the time it takes to typically implement an ML process under human supervision, but also improving the accuracy of the model when compared to those trained and deployed by humans. By enacting this, it allows organizations a pathway into ML, and freeing up the time of ML data practitioners and engineers, allowing them to focus on more-pressing and complex challenges.
There are several different applications of AutoML:
According to Gartner, more than 40 percent of data science tasks will be automated by 2020. This automation will result in the broader use of data and analytics by citizen data scientists and the increased productivity of professional data scientists. For this user group, autoML tools usually offer an easy-to-use point-and-click interface for loading data building ML models. Rather than automating an entire, specific business function, like marketing analytics or customer analytics, most autoML tools focus on model building. Yet, most autoML tools and ML platforms don’t address the issues of continuous data preparation, data selection, feature engineering, and data unification. This proves to be a challenge for citizen data scientists, who must keep up with massive volumes of streaming data and identifying non-obvious patterns. Often, they are not ready to analyze real-time streaming data. And poor business decisions and flawed analytics can arise if the data is not analyzed properly.
To automate internal processes, particularly building ML models, some companies have turned to autoML. You might know a few of them – specifically Facebook and Google. And Facebook is widely on top of ML, training and testing about 300,000 ML models each month, essentially building an ML assembly line to deal with so many models. Asimo is the name of Facebook’s autoML engineer, which automatically generates improved versions of existing models. Google is also joining the ranks by developing autoML techniques for automating the process of discovering optimization models and automating the design of machine learning models. According to Google, it is currently developing a process for machine-generated architectures.
In some cases, once ML models are built and a business problem is defined, it’s possible to automate entire business processes. It requires pre-processing of the data and appropriate feature engineering. Companies actively using autoML for the whole automation of specific business processes are Zylotech, DataRobot and Zest Finance.
Zylotech was created for the whole automation of customer analytics. The platform features a variety of automated ML models with an embedded analytics engine (EAE), automating the enter ML process for customer analytics like unification, feature engineering, discovery of non-obvious patterns, data prep, and model selection. Zylotech allows data scientists and citizen data scientists to leverage complete data nearly in real time, which enables personalized customer interactions.
DataRobot was created for the entire automation of predictive analytics. The platform automates the entire modeling lifecycle, which includes transformations, data ingestion and algorithm selection. The platform can be personalized so it can be customized for specific deployments like high-volume predictions and building a huge quantity of different models. DataRobot helps citizen data scientists and data scientists quickly apply algorithms for predictive analytics and quickly build models.
Finally, ZestFinance was created for the whole automation of specific underwriting tasks. The platform automates model training, deployment and data assimilation and explanations for compliance. It utilizes ML to analyze traditional and nontraditional credit data to score potential borrowers who may have next to no files. AutoML is utilized to provide tools for lenders to train and deploy ML models for specific use cases such as marketing and fraud prevention. It also helps financial analysts and creditors to make better risk assessments and better lending decisions.
Data may be the new oil, but even crude oil needs to be “cracked” before becoming useful molecules. Similarly, customer data must be refined before insights can be drawn from it with embedded molecules. Therefore, data is not instantly valuable, but becomes useful after it is collected, cleansed, enriched, and made analysis-ready.
The autoML approach assists businesses in using ML successfully as potential business insights are hidden in places where only ML can reach. No matter what industry you’re in, autoML is the methodology needed to extract and leverage this valuable resource.
And as organizations are increasingly depending on civilian data scientists, Gigabit Magazine predicts that 2020 is likely to be the year that enterprise adoption of autoML will go mainstream. AI and ML practices and tools will become more embedded in businesses’ everyday operations and thinking as they become more empowered to identify projects whose insight will drive better decision-making and innovation.
At Stefanini, our data science and advanced data analytics take your business to the next level. We leverage artificial intelligence and machine learning to automate your business. To learn more, visit our website or contact us.
See whats trending
Protecting Data Requires a Joint Effort from the Legal and IT Industries – Stefanini EMEA CEO Speaks at World Jurist Association Event
Our data science & advanced data analytics can take your business to the next level. Ask our experts to know more.
Hi. Need help?