In this article, we'll introduce QPython and give some examples of how your data scientists could use it for interacting with the Quantexa platform.
The Quantexa Decision Intelligence Platform is a powerful tool underpinned by a set of capabilities for performing complex analytics with Quantexa-processed data, such as resolved entities and generated networks. The powerful Assess framework aids in the creation of analytical models that can process huge volumes of data, produce alerts, and analyze data both in-batch and in real-time.
The core Assess framework is written for Scala to allow highly performant write-once scores. But the Quantexa platform also provides data scientists with a convenient interface to make their experience of working with Quantexa easier and more efficient. Since the most common language in the Data Science community is Python, we have created a QPython module that allows data scientists to interact with data in a familiar way.
Our latest module, QPython provides a pythonic interface to Quantexa to enable data scientists to work in a language that is more natural to them. The module consists of a library of functions that allows extracting entities and networks created by Quantexa, analyzing them, so that data scientists can create, train and deploy network context-based machine learning (ML) models - while keeping their work within Python.
Three QPython Use Cases
There are multiple ways Quantexa users can benefit from QPython module. Below are the three key QPython use cases:
QPython for ad hoc analytics
QPython provides a convenient way to perform ad hoc analytics. Data scientists can extract Quantexa-generated networks to Python, flatten the data, and use rich and familiar Python libraries for data interrogation. Once the network of interest has been located, using QPython 1.3, data scientists can share the information with business-side users (investigators, analysts) by sending the information to the Quantexa UI in the form of a ‘task’. Business users can then review the task, the network of interest, and the scores written in Python that highlight the insightful information or risk, or any extra data provided by the data scientist in a familiar Quantexa UI.
For example, Financial Crime experts frequently identify news events which require urgent responses. The context generated from a Quantexa network can quickly lead to the identification of clients that might be more at risk of such events. Data scientists can work with experts to hone queries, and get rapid feedback on the potential cases identified.
QPython as a development tool
QPython is also an effective tool for exploring data and testing hypotheses that can be productionalized at a later date and added to the risk detection model within Assess.
Even when the analytical model is well-designed and defined, business users often have additional questions that they want to explore in addition to existing detection models. If any newly detected risk indicators need to be added to the production model, QPython provides an alternative to Scala conversion. Users can integrate a Python score directly with Assess so that it can be considered in the alerting process. One particularly common use case of this functionality is integrating ML models that are more native to Python and expensive to convert and maintain in Scala.
QPython for tuning and improvement
Finally, QPython can be used for analyzing the outcome of scoring and assessing the parameters of the created risk detection model in Assess. Users can extract the output of scores and scorecards into Python to review the data and make decisions on changing score parameters or scorecard structure (such as weights and severities of the scores). Python tools such as Notebooks provide an effective way of recording and justifying the changes.
These are just some of the examples QPython could help data scientists interact with Quantexa-processed data. The functionality continues to evolve, making the experience of working with Quantexa more efficient and effective for data scientists.