Predictive Analytics with Strategically Missing Data.

Area of Science:

• Data Science
• Machine Learning
• Predictive Analytics

Background:

• Real-world data often has strategically missing values due to intentional concealment by data providers.
• This strategic data omission occurs in various domains like finance, admissions, and marketing, impacting decision-making.
• Existing methods struggle to address the incentive problem behind strategically missing data.

Purpose of the Study:

• To develop a novel approach for handling strategically missing data in regression prediction.
• To create a mechanism that incentivizes data providers to disclose truthful information.
• To minimize imputation errors for missing values in predictive models.

Main Methods:

• Utilizing Support Vector Regression (SVR) models to derive imputation values for missing data.
• Developing a framework that aligns data provider incentives with accurate data disclosure.
• Applying the proposed method to real-world datasets for validation.

Main Results:

• The proposed method effectively imputes strategically missing data.
• Support Vector Regression models are leveraged for accurate imputation.
• Imputation errors are minimized under specific conditions, as demonstrated by experiments.
• The approach incentivizes data providers to reveal true information, improving data quality.

Conclusions:

• The novel approach effectively addresses strategically missing data problems in predictive analytics.
• Support Vector Regression provides a robust foundation for imputing missing values.
• The method offers a practical solution for decision-makers facing data concealment.
• Experimental validation confirms the approach's effectiveness on real-world data.