Loading...
office@audascience.com +972-77-8011500

AuDaScience Profile Builder BRAINs for Marketing Machine Learning Modeling

Collect batch and streaming data from DWH, digital sources and legacy systems and AUTOMATICALLY construct and update, in real time, an aggregated Customer Data Platform for a real 360° customer view.

Every data scientist knows that ~80% of data science work required for developing and implementing machine learning (ML) models for marketing (such as cross-sell prediction, churn prediction, lifetime value prediction), is data preparation work.

The algorithms used in the machine learning model are important but the data that goes into the modeling process is key to really improving the prediction level of the model.
This is the reason why data scientists invest the majority of time spent on the development of ML models in creating aggregated, Customer-level attributes and building calculated attributes – such as trend over time and ratios attributes. All for purposes of adding attributes that might potentially have a significant positive statistical impact on the model's prediction.

Unfortunately, data preparation for modeling takes time and requires massive amounts of manual work, even when using ETL tools. Data preparation for updating aggregated data in real time, for the purpose of real time scoring, is even more complex and requires vast experience in RT programming on big data infrastructure.
AuDaScience Profile Builder BRAINs enables the automatic development and update of an aggregated Customer Profile table using engines:

  1. That runs on Spark for creating and updating an aggregated Customer Profile table from raw historic data, on a periodic basis (i.e. daily, monthly, ...).
  2. That runs on Cassandra for creating and updating an aggregated Customer Profile table from streaming data, in real time.

Profile Builder BRAINs enables an analyst to :

  1. Collect customers' raw data (batch or streaming data from digital assets, such as website or mobile app, or transactional systems)
  2. Define the set of aggregations required to run on each raw data (i.e. the maximum amount that a user has entered in the loan amount box on the loans page on the website during the past week).
  3. Using Profile Builder BRAINs engines transforms this logic into a deployment process which runs on batch or RT on the organization's big data & open source environment.

The Customer Profile is updated in real time and enables real time triggering and scoring of machine learning models developed automatically using Sales BRAINs.
In addition, the RT-updated Customer Profile can be exported on a periodic basis (every hour, day, ...) into the data lake or DWH environment for the ongoing use of the organization's analysts.

The main advantages of Profile Builder BRAINs are:

  • Saves time on data preparation work that could take many man-months using coding or ETL and shorten time-to-results from months to hours.
  • Integrated with AuDaScience BRAINs applications for real time scoring of marketing machine learning models, enabling fully automated data science and ML processing from data preparation, through ML modeling and real time scoring, without integration work.
  • Running on the organization's big data and open source environment, assisting in monetizing investment already effected on the establishment of big data environment.