Advancing Artificial Intelligence
& Machine Learning
AI, Meta Data Decision Making
Artificial Intelligence brings significant impacts on current business. Today, it is not just a computer which you used for lab but a resource which is everywhere to solve real-life business problems. AI becomes one of the important competitive advantages which creates new values and enables company to grow.
Machine learning is an area which computers can do self-learning without clear programming.
ex. Autonomous Vehicle, Spam filer
Depends on Data training types, Machine learning can be classified as below
An action which prediction model learns from past-data to predict the right answer
Target information based general predictive modeling
Financial Institutions credit prediction : Delinquent characteristics checking machine learning model development
Analysis to find similar group or pattern with no target
Customer Segmentation Plan by similarity grouping, clustering and linkage analysis
Classification by customers’ behavior : Customer research to create various segments by clustering analysis
Algorithm learns to react
To an environment
Method which to find the optimum strategy from learning by trial and error
Draw optimum strategy by reinforcing from measuring performance
Google’s Self-driving car and Limit optimization and etc.
‘Next best offer’ model development for Call-center group
How Predictive Analytics and Machine Learning are related?
Development Tool: SAS, Model Builder, Model Development Studio, etc.
At development stage, Artificial Intelligence can be applied by machine learning algorithm and retraining automation process
PANIROIS provides data analytic service with machine learning methodology
What is the major supervised learning algorithm for credit assessment model?
The application of collective intelligence
•Developing various models randomly and predicting a target by combining the developed models
•High predictive ability due to various models’ outcomes based on the majority decision or collective intelligence principle
Gradient Boosting Machine
Overcoming of weak prediction models
•Improvement in discrimination by adding relatively weak prediction models to training data
•High discrimination ability due to the growth through overcoming weaknesses of the model
Maximizing the prediction with
a nervous system principle
•Predicts a target by applying complex neural network functions like human brain
•Possible to predict the target with complex and various patterns through elaborate predictions