Research libraries
Data management, cleaning, and analysis
The internet is full of data science and statistical learning libraries, most of them open source and well documented: it's possible to quickly find a library that implements some specific operation quickly and (most likely) correctly. Along with these benefits, there are also drawbacks:
-
The user needs to know exactly which technique/tool to use and which implementation of it is appropriate and cobble together different tools from different sources
-
In a production environment, managing many open-source libraries is risky and challenging
-
APIs and functionality change from time to time, potentially (silently) breaking dependent code
-
Regularly reviewing and conforming to each library's license terms adds compliance burden.
tpalib is solves a number of these problems by providing
-
a single license
-
a curated set of tools for predictive modelling problems that work together
-
managed updates and API consistency
-
responsive support
Custom Tools
Model and data visualization
An amazing model is useless if there's no way to use it and a perfect forecast is worthless if the decision makers can't see and understand the results. We believe that as much time should be spend understanding a model and its results as researching and developing it in the first place.
To help with all phases of a data/predictive modelling project, we can create:
-
Dashboards: Visualize data and results historically and as they evolve in the future
-
Simulators: Understand model output, run what-if scenarios
-
Data collection tool: gather data from users, websites, documents in a structured and easy to use way