Innocent Internet users are paying the price for the increasing prevalence of fraudulent websites, which generate billions of dollars in illegal revenue. There is a need for intelligent technologies to recognise harmful websites as online criminal activity rises. It has been demonstrated that URL analysis is a useful method for identifying phishing, malware, benign, and defacement. For URL categorization, previous studies have used lexical aspects, network traffic, hosting data, and other techniques. These methods necessitate time-consuming searches that cause real-time systems to experience severe delays. This paper represents a simple method for classifying dangerous websites that relies just on lexical URL analysis. To examine the accuracy of web URLs, machine learning models like Random Forest Classifier, XG Boost, and Light Gradient Boosting Machine are utilised.