Emotions serve as a means of perceiving the affective states of others, but the comprehension of sarcasm requires additional cognitive exertion. Sarcasm, a kind of linguistic irony, involves the use of words or phrases that diverge from their literal meaning. Academic researchers continue to dedicate their efforts towards the development of an algorithm capable of accurately detecting and identifying instances of sarcasm. The challenge of teaching a computer to perceive sarcasm is not a straightforward undertaking, since people even frequently struggle to comprehend it. The demand for Hybrid Machine Learning (HML) is seeing a significant increase in the context of detection and classification tasks. This study examines three distinct categories of characteristics: lexical, sarcastic, and contextual factors. The aforementioned feature sets are used in the task of classifying tweets into two categories: sarcastic and non-sarcastic. This study introduces a feature set characterized by a sardonic tone, along with a hybrid machine learning technique that demonstrates enhanced accuracy. The experimental findings demonstrate that the suggested hybrid machine learning approach attains a 95.30% accuracy rate when applied to sarcastic feature sets. This performance surpasses that of established machine learning methods such as k-nearest neighbor, random forest, support vector machine, and decision tree in terms of feature selection.