Descriptive Analytics Framework in Artificial Intelligence
Areteans Descriptive Analytics Framework in Artificial Intelligence, powered by Inferential Statistics, is a prototype solution to summarize, visualize, and represent the historical discoveries of a Pega customer dataset that is derived from a sample or population. The solution occurs during the feature engineering and preprocessing phase of the data science project life-cycle. This framework will help explore statistical information and pattern recognition from customer behavioral, demographic, or transactional past data and to build a state-of-the-art AI model.
- Descriptive Analytics helps data scientists enable data visualization before configuring your predictor and model selection.
- Descriptive Analytics is a powerful aid to provide statistical information about numerical and categorical variables in a dataset and to highlight potential relationships between explanatory variables with response variables.
- The framework provides an AI Heatmap Plot of data to check the statistical interrelation among explanatory variables with response variables - where the intensity of the color represents the depth of the relationship.
- Framework generates Box and Whisker Analysis for graphically representing dispersion and variation in a set of data with eight different KPIs.
- Histogram plot of predictors is to check spread, imbalance skew, etc. of data to easily observe the pattern of distribution, i.e., Gaussian-Normal Distribution, Exponential Distribution, Bi-model Distribution, etc.
- Statistical data preprocessing analytics for feature engineering with detailed analysis of mean, count/frequency of data, standard deviation, minimum and maximum value 25-50-75 percentile, etc. of numerical predictors.
- Detect class imbalance of the target class (for classification type of problem) to take appropriate transformations or measures in removing bias of target.
- Encapsulate the characteristics of a sample from population, such as a predictor's mean, median, standard deviation, or dispersion of data.
- Determine the pattern of the data, help eliminate the presence of noise (e.g. outlier detection) - all before the AI model develops.
- Suitable data processing and data clean up (processing of nearly constant data, blank-NA-missing data, outlier data, imbalance data, redundant data, etc.) is done with the knowledge of descriptive analytics to analyze the behaviors of predictors by multiple techniques before importing the dataset for the Prediction Algorithm treatment.
- Pega data scientists can pre-process on a unified platform of Pega without using additional external tool/coding platforms.
- Business can visualize important statistical KPIs and metrics of historical performances from raw data before making the decision, with the help of the solution dashboard.
- Data engineers can recognize the patterns of trends and anomalies which are hidden in raw data on the click of a button within Pega.
- Users can take advantage of the plugged-in Pega Customer Decision HubTM Portal.
- Help to prepare, extract, and generate relevant date for modeling in Pega.
- Robust and scalable to support future enhancements with customer requirements/datasets.
- Industry: Technology Services
- Product Capability: Low-Code App Development
- Pega Version compatibility: 8.1, 8.2, 8.3, 8.4, 8.5, 8.6, 8.7
Offering TypePackaged Service Offering
Published OnFebruary 23, 2023