By: Luba Gloukhova, Founding Chair, Deep Learning World
In anticipation of his upcoming conference presentation at Deep Learning World Las Vegas, June 16-20, 2019, we asked Alex Glushkovsky, Principal Data Scientist at BMO Financial Group, a few questions about his work in deep learning. Catch a glimpse of his presentation, Robotic Modeler for Marketing Campaigns, and see what’s in store at the DLW conference in Las Vegas.
Q: In your work with deep learning, what do you model (i.e., what is the dependent variable, the behavior or outcome your models predict)?
A: Analytically supporting acquisition, attrition, sells (cross-sells, up-sells), or customer stimulations, the focus is on both immediate responses but, most importantly, on sustainable long term microeconomic results. It means finding win-win equilibriums of customer-business relationships. The challenge is to predict customer behavior using past examples which by nature are rather sporadic, and not stable reproducible patterns, in a very noisy and dynamically changing competitive environment.
Q: How does deep learning deliver value at your organization – what is one specific way in which model outputs actively drive decisions or operations?
A: Applying predictive modeling for marketing campaigns, it delivers multiple cross-leveraged effects of (1) increased revenue by more adequate offerings, (2) cost saving by decreasing irrelevant targeting, (3) greater customer experiences with more favorable and less annoying customer communications, and (4) unprecedented operational efficiency by decongesting bottlenecks and by shortening lead times while providing exceptional accuracy.
Q: Can you describe a quantitative result, such as the performance of your model or the ROI of the model deployment initiative?
A: Integrating the automated data preparation layer with an automated modeling engine leads to outstanding robotic modeler performance that I will be discussing at upcoming Deep Learning World. It allows for crunching tens, if not hundreds, of marketing campaigns that run monthly and automatically train, validate, and select models within hours if not minutes.
Q: What surprising discovery or insight have you unearthed in your data?
A: It is the power of discovering outliers. Keeping in mind the confidence of outcomes, it is desirable to have results representing the “bulk” part of data. However, our world is tremendously varying with uncommon or unexpected elements. Businesses should not ignore outliers, but look for them instead. It will help discover new insights and will point to possible improvements.I am also very amazed by modern analytical abilities to assemble meaningful business “jigsaw puzzles” by having millions of data pieces with usually loose “cuts of edges”.
Q: What excites you most about the field of deep learning today?
A: It is a great professional community with different backgrounds, open data sources for experiments, outstanding intelligence of available methodologies and algorithms, and … infinite opportunities of possible practical applications and analytical approaches.
On the other end, there is a gap between achieved great results and developed fundamentals of deep learning. Isolated deep learning elements are sensible and well defined, such as a loss function, backpropagation, or convolution, but specific architecture of a deep learning application is still an art rather than a science where empirical tuning still plays a significant role. Today, we do not have a magic formula on how the optimal deep learning solution should look like even having standardized data. I expect a lot of discoveries on principles of deep learning in the future.
Q: Sneak preview: Please tell us a take-away that you will provide during your talk at Deep Learning World.
A: It can be observed that exceptional results of deep learning have been achieved in areas where there are both standardized objectives and standardized data. We usually do not have standardized data in business, but creating common data sources of targets and pre-engineered features suitable for specific business areas will allow to build automated modeling frameworks, such as a robotic modeler for marketing campaigns that I will be discussing at Deep Learning World. Integrating the automated data preparation layer with an automated modeling engine, it allows for scrolling of executed marketing campaigns, creating modeling datasets, fitting competing models, validating results, selecting the best models based on defined criterion, and generating scoring codes and documentations – all in an automated way. Furthermore, the robotic modeler framework supports generalized deep learning assembling business targets and features. Systematically running the robotic modeler provides additional benefits such as perceiving input feature importance from various campaigns or estimating cross-campaign effects. It empowers “hyper-learning” derived from campaign modeling. Overall, the framework provides exceptional operational efficiency, benefits customers and organization, and shifts data scientist efforts from routine manual tasks of extracting data and crafting modeling processes for each marketing campaign to more intelligent work of enhancing input features and tuning hyperparameters across multiple campaigns.
As a sneak preview one can read my short article here in “The Predictive Analytics Times”.
—————————–
Don’t miss Alex’s presentation, Robotic Modeler for Marketing Campaigns, at DLW on Wednesday, June 19, 2019 from 3:55 to 4:15 PM. Click here to register for attendance.
By: Luba Gloukhova, Founding Chair, Deep Learning World
This content is restricted to site members. If you are an existing user, please log in on the right (desktop) or below (mobile). If not, register today and gain free access to original content and industry news. See the details here.
 |
 |
 |
