<p>In this article, we proposed a generalized data-driven class of exponential product-type imputation methods for handling missing data in survey sampling, and the corresponding point estimators are developed. We then derive the bias and mean square error expressions of the proposed estimators. The proposed estimators are more efficient estimators as compared to the existing estimators considered. A numerical illustration is carried out by considering varying population sample sizes ranging from 30 to 40% and response rates between 60 and 90% for two different real data sets. Furthermore, a simulation study is conducted under controlled negative correlation structures using the same datasets as in the numerical illustration, with correlation levels set at − 0.80, − 0.85, and − 0.95, while similarly varying the sample size and response rate considered in the numerical illustration. The results from both numerical and simulation analyses show that the proposed estimators have higher efficiency as compared to existing estimators.</p>