This paper introduces two improved almost unbiased estimator for estimating the finite population mean in neutrosophic settings, incorporating two auxiliary variables to handle indeterminate data more effectively. We have improved the classic ratio and product estimator with new estimators (t_hN) and (t_h1N), offering enhanced accuracy when applied to uncertain real-life data. Through theoretical derivations and empirical validation using agricultural data (rice yield with climatic variables), we demonstrate that our estimators perform better in term of both accuracy and efficiency. The results show significantly higher Percentage Relative Efficiency (PRE) and lower Mean Squared Error (MSE), highlighting the method’s effectiveness for scenarios involving imprecise or indeterminate data. This study develops a framework for better statistical estimation by merging neutrosophic logic with classical sampling methods to handle imprecise data effectively.