List of Features in Dataset.

Choosing the appropriate career path poses a significant hurdle for students, especially when time is constrained. This research addresses the challenge of career prediction by introducing a method that integrates additional attributes, refines feature prioritization, and streamlines feature selection to enhance prediction precision. The key objectives of this study are to pinpoint pertinent features, accurately rank them, and enhance prediction accuracy by eliminating non-essential features. To accomplish these aims, three methodologies are employed: Feature Fusion and Normalization (FFN) for precise data identification, Average Feature Ranking (AFR) utilizing a blend of Random Forest (RF) and Linear Regression (LR) for feature prioritization, and Improved Prediction with Weighted Characteristics (PWF) which integrates Principal Component (PC) analysis for feature reduction. The prediction performance is assessed using a hybrid Multilayer Perceptron (MLP) classifier with 5-fold cross-validation. The outcomes reveal that the hybrid approach yields a superior feature set for prediction. The top twelve ranked features are determined by averaging each feature’s RF scores and coefficients. The achieved accuracy (ACC), precision (P), recall (R), and F1 scores stand at 87%, 87%, 86%, and 86%, respectively, with an Area Under the Receiver Operating Characteristic Curve (AUC-ROC) value of 92%. These findings underscore the efficacy of the proposed hybrid learning technique in accurately forecasting career trajectories.

在时间紧迫的情境下，选择适配的职业道路对学生而言是一项重大挑战。本研究针对职业预测领域的这一难题，提出了一种融合额外属性、优化特征优先级排序并精简特征选择流程的方法，以提升预测精度。本研究的核心目标在于精准识别相关特征、对其进行准确排序，并通过剔除非必要特征以提升预测准确率。为实现上述目标，本研究采用了三种方法：用于精准数据识别的特征融合与归一化（Feature Fusion and Normalization, FFN）、结合随机森林（Random Forest, RF）与线性回归（Linear Regression, LR）实现特征优先级排序的平均特征排序（Average Feature Ranking, AFR），以及融入主成分（Principal Component, PC）分析完成特征降维的加权特征改进预测（Improved Prediction with Weighted Characteristics, PWF）。本研究采用混合多层感知器（Multilayer Perceptron, MLP）分类器结合5折交叉验证对预测性能进行评估。实验结果表明，该混合方法可生成适用于预测的优质特征集。研究通过对每个特征的RF得分与系数取平均值，确定了排名前12的特征。最终获得的准确率（ACC）、精确率（P）、召回率（R）与F1值分别为87%、87%、86%与86%，受试者工作特征曲线下面积（Area Under the Receiver Operating Characteristic Curve, AUC-ROC）达92%。上述结果证实，所提出的混合学习技术可有效精准预测职业发展轨迹。