Abstract
Consumer confidence surveys are key indicators of households' economic expectations and support monetary and policy decisions. However, prior studies mainly rely on traditional statistical methods, leaving a research gap in capturing the complex factors shaping public perception. This study employs machine learning techniques to uncover patterns in consumer sentiment regarding India's general economic conditions over a one-year horizon. Data collected from Reserve Bank of India's Consumer Confidence Survey from Jan 2022 to June 2024. Three were 90,000 responses with 28 variables. The data classified participants' perceptions of future economic conditions into three imbalanced classes: worsened, improved, and remained the same. Machine learning algorithms; Decision Tree, Random Forest, Gradient boosted algorithms, Multilayer Perceptron and Probabilistic neural network were chosen from the literature review. This study compared machine learning models' performance before and after treating imbalanced class problems. The model's classification accuracy had improved when we address imbalanced class problem using SMOTE. Among all classifiers, Random Forest algorithm outperformed the rest with an F1-score of 0.77. In addition, the general profile of respondents with different perceptions of India's general economic conditions were identified. This data driven insights could be helpful to the policymakers to design economic policies to cater various consumers.