<![CDATA[This research aims to develop a predictive model for tidal inundation at Tanjung Perak Port in Surabaya, a region identified as critical and highly susceptible to such events. The foundational data incorporated comprises hydrometeorological indicators, such as lunar cycles, tidal patterns, and precipitation levels, which were sourced from the BMKG Tanjung Perak Maritime Meteorological Station. A dataset comprising 26,275 individual data points was compiled and subsequently partitioned into training sets (80% of the data) and validation sets (20%) via randomization. This apportionment is intended to support the robustness and applicability of the developed model. The initial data preparation phase involved techniques such as data normalization, imputation of missing values, and the determination of variable weights based on their respective degrees of impact. Subsequently, two distinct machine learning methodologies were employed to construct the predictive framework: Gradient Boosting (specifically, XGBoost) and Logistic Regression. The efficacy of the resultant models was rigorously assessed using various metrics, including accuracy, confusion matrix analysis, ROC-AUC scores, and feature significance analysis. Analysis of the outcomes indicated that the Gradient Boosting model achieved a superior accuracy of 99.96%, whereas Logistic Regression attained 99.85%. An examination of the features revealed that lunar cycles and tidal conditions were the principal determinants of tidal inundation, with precipitation exerting a comparatively minor effect. These observations substantiate the efficacy of integrating suitable data preparation techniques with machine learning methodologies to achieve precise predictive outcomes. The principal contribution of this investigation is the establishment of a computational framework to facilitate the development of an advanced warning system for tidal flooding, thereby aiding hazard reduction and limiting adverse societal, financial, and operational consequences in littoral regions.]]>
