Predictive Analytics
Concepts and Applications
INTRODUCTION
Predictive Analytics is a data driven process to forecast what may happen in future.
It makes use of historical data and methodologies from statistical practices or machine learning techniques (tree-based algorithms, deep learning) to develop predictive models.
HOW DOES PREDICTIVE ANALYTICS WORK?
Data analytics plays a crucial part by identifying the correct metrices for a given problem and cleaning and processing the relevant data fields.
A mathematical model is used to learn the relationship of various fields from training data
The patterns in the form of a model are then used to make predictions about new data.
TYPES OF PREDICTIVE MODELS
SOME APPLICATIONS OF PREDICTIVE ANALYTIC
Marketing
Customer behavior, Offer right products and discounts
Retail
Customer loyalty, Churn analysis, forecast trend and revenue
Finance
Default prediction, fraud detection
Healthcare
Susceptibility to diseases like heart disease, diabetes etc.
Manufacturing
Quality control
CHALLENGES OF PREDICTIVE ANALYTICS
- Availability of right data for a particular problem from data overload
- Incomplete and inaccurate data incorrectly influence model training
- Team requires people with domain knowledge and technical skill
- Automated system can unnerve online customers by predicting their intent and influencing behaviour
- Get people to adopt the new model primarily due to distrust in the metrics used and reluctance to leave the comfort zone of using old solutions
WAY FORWARD
- Data volume from various sources is increasing at a rapid rate providing valuable resource for model building
- Increased advancement of AI and ML has positive impact on accuracy of models
- Rise of explainable AI helps in transparency and explainability
- Real time predictive analytics analyzes data as it is generated, thus helps in making instant decision in the face of changing marketing conditions