Implementation of web product search engine : parallel incremental web crawler

One of the main objectives in designing a Parallel Incremental Web Crawler is to provide a solution to the problem of designing a large scale web-based Content Based Image Retrieval (CBIR) system. Our CBIR system has indexed more than 1 million images crawled from various Business to Consumer (B2C) websites till date. The Internet traffic today is getting more complicated and analyzing how websites are interlinked and their content similarity is important for Web Mining. Due to the growing and dynamic nature of the web, it has poses unprecedented scaling challenges to traverse all URLs in the web documents and handle these URLs, so it has become imperative to parallelize a crawling process for extraction of useful data from the web. In this report, we have proposed a novel architecture of a parallel crawler with an optimization model which is scalable and resilient against system crashes while maximizing the download rate and minimizing the overhead from parallelization based on API and domain specific crawling. We will also discuss how our crawling module is realized to make crawling task more effective and scalable in the collection process of data retrieval without recursive crawling on the same honey pot. We will also be discussing on the storage of extracted data using certain data management techniques and also image processing techniques such as Spatial Anti-Aliasing and enhancing by the crawler when an image is being processed and stored. Finally, several experiments were conducted to evaluate the processed data quality as well as the effectiveness of the algorithms parallel performance in the web crawler. In the experiment, several benchmarking test was also conducted to evaluate the CPU resource utilization as well as the freshness of the eVISE operational database.Bachelor of Engineering (Computer Science