International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 03 | Mar 2025
p-ISSN: 2395-0072
www.irjet.net
Drug Classification Using Selective Naive Bayes Mutual Probabilities on features K.JayaLakshmi1, Dr.K.Venkataramana2 1Student, Dept of MCA , KMMIPS, Tirupati
2Professor , Dept of MCA , KMMIPS, Tirupati
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract- Selective Naive Bayes (SNB) is an enhancement
tasks. However, the traditional Naive Bayes approach assumes that all features contribute equally and independently to the classification process, which often leads to inaccuracies when dealing with irrelevant or redundant attributes.
of the traditional Naive Bayes classifier that incorporates feature selection techniques to improve classification accuracy. The standard Naive Bayes algorithm assumes feature independence, which may lead to misclassification when dealing with irrelevant or redundant attributes. SNB addresses this limitation by selecting only the most informative features, thereby optimizing the model’s performance. This paper explores the methodology of SNB, including various feature selection metrics such as mutual information, chi-square, and information gain, which help refine the classifier's input variables. The study also evaluates the impact of feature selection on classification accuracy and computational efficiency across multiple datasets. Applications of SNB in domains such as text classification, spam filtering, medical diagnosis, and sentiment analysis are discussed, demonstrating its effectiveness in real-world scenarios. A comparative analysis with traditional Naive Bayes models highlights the improvements in precision, recall, and overall classification performance achieved through feature selection. Additionally, the paper examines the challenges associated with SNB, including the trade-off between feature reduction and information loss. The results of this study emphasize the importance of feature selection in probabilistic classification and validate SNB as a powerful alternative to the conventional Naive Bayes approach. This research contributes to the ongoing advancements in machine learning by encouraging the development of more refined probabilistic models that balance accuracy and efficiency. Future work in this area could further optimize SNB through hybrid feature selection methods and adaptive learning techniques.
To address this limitation, Selective Naive Bayes (SNB) introduces a feature selection mechanism that identifies and retains only the most relevant attributes for classification. By eliminating non-informative features, SNB enhances model accuracy, reduces computational complexity, and improves generalization to unseen data. Various feature selection techniques, such as mutual information, chi-square analysis, and information gain, are used to optimize the performance of SNB. This paper provides an in-depth exploration of Selective Naive Bayes, analyzing its methodology, advantages, and applications. We compare SNB with the traditional Naive Bayes classifier to demonstrate its effectiveness in realworld scenarios such as text classification, spam detection, sentiment analysis, and medical diagnosis. Additionally, we discuss the challenges and potential improvements in SNB, including the integration of hybrid feature selection methods and adaptive learning strategies. The remainder of this paper is organized as follows: Section II reviews related work, Section III presents the methodology of SNB, Section IV discusses experimental results, and Section V concludes with future directions.
2. LITERATURE SURVEY ON SELECTIVE NAIVE BAYES (SNB)
Key Words: Selective Naive Bayes, Feature Selection, Machine Learning, Probabilistic Classifier, Text Classification, Data Mining, Classification Algorithm, Supervised Learning.
Selective Naive Bayes (SNB) is an extension of the traditional Naive Bayes classifier that incorporates feature selection techniques to enhance accuracy and computational efficiency. Several researchers have explored SNB across domains such as text classification, medical diagnosis, cybersecurity, and financial fraud detection.
1. INTRODUCTION Machine learning has become an essential tool in data classification, with numerous algorithms designed to enhance accuracy and efficiency. Among these, the Naive Bayes classifier is widely used due to its simplicity, scalability, and effectiveness in probabilistic classification
© 2025, IRJET
|
Impact Factor value: 8.315
John & Langley (1995) [1] were among the first to analyze the impact of feature selection in Naive Bayes, concluding that removing irrelevant attributes improves classification accuracy. McCallum & Nigam (1998) [2] studied SNB in text
|
ISO 9001:2008 Certified Journal
|
Page 821