5. METHODOLOGY The following methods have been implemented in this document:5.1. Data Record MiningData records are nothing more than the maximum amount of information present on the web in the form of structured objects. These data records are important because they represent the necessary information of the host pages, for example, different lists of products and services [15]. The proposed methodology avoids all the drawbacks of the existing system. Record extraction finds QRRs from the web page and performs the following two steps: data region detection and actual segmentation. Figure 16 shows the QRR extraction structure and cluster formation. This method identifies QRRs from the web page. Based on the users query, QRR can be generated. The tag tree will be formed from the QRRs. Tag tree construction builds a tag tree for the page where the root nodes begin. The HTML page represents all tag nodes including child tags. Using data extraction based on partial tree alignment, i.e. the DEPTA algorithm, a tag tree can be generated [2]. DEPTA, also known as Mining Data Records-2, is a more efficient and effective algorithm than the Mining Data Records algorithm. This algorithm segments the page to detect each data record without extracting data elements from it. The main task is how to match equivalent data items or fields from all data records [16]. The distance string editing algorithm can also be used for similarity identification [17] between records. The following four sections detail QRR extraction:5.1.1. Data Region Detection Detects all probable data regions, which usually contain vigorously generated data, in a top-down manner starting from the root node. The data region detection algorithm detects data regions in a top-down manner. Start from the root node of the query result pa...... middle of the sheet ......ame x neuron value. This is shown in Figure 27.5.5. SOM Algorithm The Self Organizing Map is an unsupervised algorithm that can be used to cluster [7] M patterns across n classes. Algorithm1. Initialize the weight vector wij. Random values ​​can be assumed.2. Consider an input vector.3. Traverse each node of the map.a. Calculate the Euclidean distance to find the similarity between the input vector and the weight vector.b. Find the winning vector for the minimum Euclidean distance.4. Calculate the new weights by bringing them closer to the input vector. W v (t+1) = W v (t) + ϴ (t) α (t) (D (t) – W v (t))5. Update the learning rate α.6. EndOther clustering techniques [8] divide M number of patterns into n predefined classes. We use the Self Organizing Map algorithm to divide different types of input models in the form of web pages into different classes.