Imputing Data Gaps in Economic Surveys Using Fuzzy Sets and Artificial Intelligence Technique

Purpose: This paper develops a novel approach to impute data gaps in economic surveys. In contrast to classical methods relying on statistical analysis of survey data, more advanced prediction techniques combined with fuzzy sets are applied to effectively address missing data. Design/Methodology/Approach: The paper proposes an unconventional approach that integrates advanced prediction methods with fuzzy sets for imputing missing data. The effectiveness of the method is tested on the extensive dataset from the Polish Panel Survey (POLPAN), which was conducted every five years from 1988 to 2018. The survey contains a wide range of questions asked over successive waves, enabling a comprehensive analysis of the method for imputing data gaps. Findings: The results of numerical experiments show that the proposed method performs highly effectively, regardless of the proportion of observations assigned to the training set. Some methods, such as Support Vector Machine (SVM), did not prove suitable for imputing this dataset. The choice and number of explanatory variables play a crucial role in the method's effectiveness, with cases where a single variable was sufficient for accurate imputation. Practical Implications: The proposed method offers practical applications for improving data quality in economic surveys, especially in large-scale longitudinal surveys like POLPAN. It provides new insights into handling missing data and optimizing the selection of explanatory variables, which can enhance the robustness of imputation techniques in complex surveys. Originality/Value: This paper contributes an original and valuable approach by combining advanced prediction techniques with fuzzy sets, providing a highly effective tool for imputing missing data. This unconventional method offers new avenues for further research in economic surveys and beyond.