Which kinds of machine learning models can I build using Pecan?
Pecan’s Predictive GenAI allows you to build a predictive model to generate predictions that address nearly any business use case, as long as you have the relevant data. Our Predictive Chat and Predictive Notebook guide users to define the right model and training dataset to fit their business needs. In more technical terms, Pecan can build binary and multiclass classification models and regression models, which can address a wide range of business questions.
How long does it take to build models with Pecan?
Not long at all! Models can be ready to deploy in hours or days, instead of the weeks or months typically required in traditional machine-learning projects. By starting with our Predictive GenAI capabilities, users rapidly arrive at a well-defined predictive question and a suitable training dataset. Our pre-built integrations make it easy to incorporate data from a variety of sources. While you can spend more time fine-tuning if you want, you can also quickly iterate and experiment, thanks to our fast modeling capabilities that prepare and train models in 15 minutes or less. Most users who can use one of our existing data integrations should arrive at a first finished model in an hour.
Can I trust automated predictive analytics?
You sure can. Automated approaches to predictive modeling are now widely accepted and used by even the most sophisticated data scientists, who rely daily on AutoML tools to quickly build, evaluate, and select the right predictive algorithms. We’ve tried and tested our automated methods for data cleansing, feature engineering, and model building with thousands of models for hundreds of customers. We’re confident in the results, and you can be, too.
How does Pecan handle automated data preparation and feature engineering?
Pecan’s automated processes detect common issues in data quality — such as missing data, outliers, or duplicated data — and fix them. The platform also looks for relationships in your data to identify new features to build, such as aggregations, trends, and other key data points. From hundreds or even thousands of potential features, Pecan identifies the features that really matter for accurate predictions (spoiler: all that is done automatically, too!). These automated steps eliminate much of the tedious, time-consuming work required to create an AI-ready dataset. This preparation also improves models’ performance and helps you get the most value from your data.
How does Pecan choose the best machine learning algorithm?
Pecan first automatically trains hundreds of predictive models using a variety of algorithms known to perform well on the kinds of data businesses usually have. Then, the models are all tested and compared behind the scenes. Pecan has the unique capability of allowing analysts to control the business metric that will be used to choose the model so that it answers your business needs. The best-performing model is selected for you to examine and refine in the Pecan interface and to deploy when you’re ready.