Trust is the ability to have confidence in someone or depend on his/her word, person, or team. Thus, if any personality is labeled as having trust problems, then such a personality is seen as having dysfunctional trust for people and experiencing danger in interacting with others. The COVID-19 pandemic created tension, stress, and threats that affected interpersonal beliefs negatively. The conflict over interpersonal trust causes disturbances in individual and societal welfare. Hence, the only way is to work towards solving the trust issue to restore trust. In this article, the authors presented an approach that involved the use of a machine learning model in line with a real-life dataset to determine cases of problems relating to trust issues swiftly and accurately. In this research, an attempt has been made to construct a model with less number or features and this in turn reduces the computational requirement. To build the proposed model a four-stage method that entails data collection through face-to-face interviews, feature selection, various classification algorithms, and the comparison of the algorithms’ performance was used. The dataset contains 91 independent variables and one dependent variable in the study. To perform feature selection several algorithms such as Information Gain, One-R, and Relief-F were used. These algorithms remove features from the lowest ranked of the dataset to higher ranked features in a recursive manner. Subsequently, the reduced dataset for training and testing of the classification algorithm was applied. The classification algorithms applied are Random Forest (RF), Random Tree (RT), Logistic Regression (LR), Multilayer Perceptron (MLP), and Support Vector Machine (SVM). The RF algorithm, with five features, gives 100% accuracy. To evaluate the performance, the authors employ several train-test split approaches and 10-fold cross-validation. The authors apply different statistical measures to assess the effectiveness of the classification algorithms. Last but not least; the suggested model has 100% accuracy in all classification matrices on a reduced feature set.