Some clients require custom built machine learning models for a unique competitive advantage. Purple Leaf Communications can help from the ground up by following an improved methodology based on CRISP-DM, CRoss Industry Standard Process for Data Mining.

We start by defining the project with clear business objectives, identifying current and new data sources, clean and prepare the data in a format required by the various models, select the best models to be trained, evaluate each with predetermined performance criteria, test, and finally operationalize for automated use.  These include models for facial and image recognition, text mining and analysis, speech recognition and text-to-speech, or traditional machine learning solutions of classification, clustering, and forecasting. In some cases, where a cloud-based model already exists, we can streamline development and deployment by configuring infrastructure to the service provider’s (AWS, Google, etc) cloud platform.