Slashgear eBay has been continuously collecting consumer data and using machine learning methods to attract more consumers and increase transaction credibility. In order to enhance the consumer's interactive experience on the site, eBay added machine learning methods to the site. For the past four years, eBay has been collecting consumer interaction data, such as search data and search clicks, and then loading this data into its machine learning system. In an interview with InformationWeek, Dan Fain, vice president of engineering at eBay, disclosed the company’s plan and business motive behind it. This is a case worthy of IT engineers who can add machine learning systems to this type of consumer-oriented application, thereby increasing the company's market share. Fain said that consumers' search habits, what pages they like to see, their favorite languages, recommended products, and image analysis are all key elements of the machine learning system, and eBay has taken this point. The most important thing is that the purpose of applying machine learning is to strengthen the search function of eBay website. Fain said in the interview: "We have a variety of machine learning models, just to ensure that the best search results for consumers." There are more than one billion items sold on eBay, and accidentally searching for results can be erroneous. Because these search results largely depend on the keywords entered by the consumer, the hyphenation is different and the results may not be the same. For example, when searching for a "sewing machine", what the consumer wants may be a sewing tool. However, the result of machine learning shows us that the sewing machine searched by consumers may be an old-fashioned sewing machine for ancestors, or it may be an antique sewing machine for collectors. The search keyword “sewing machine†triggers eBay to search for its inventory of sewing machines, while the machine learning system guesses the type of sewing machine it wants from the search history of the searcher. In addition, the machine learning system will also monitor all the consumers on the site searching for sewing machines, and observe that they are watching and leaving. Still fancy new style? How do collectors search antique sewing machines? Another example: If we search for "Swiss watches, leather straps," this is a different challenge for the site's machine learning system. General search engines return a long list of Swiss watches and a long line of leather bracelets. And if the customer wants a "Swiss watch with a leather strap," the search results obviously have no meaning. Fain said: "In the face of this search keyword, two different categories of products need to be included in each other." Despite the frequency of such search keywords is very small, the system will still consider the long tail effect when matching the best search results. Fain also pointed out that the system can link some names that are hardly to appear at the same time by analyzing consumer search history. eBay's search engine is very cautious in determining the type of search keywords, and the system will combine customer's other clues based on the content in the keyword category, and finally search out the products that consumers are interested in. Therefore, in the above search case, eBay's search engine can present consumers with hot-selling products in Swiss watches with leather straps. Everything for trading Fain said that by the end of 2015, eBay had 162 million active users. By August 2016, this figure has risen to 164 million. For eBay companies that have seen their earnings stagnate or even decline, a small increase in active users is a good news. Because the competition in the e-commerce industry is very fierce, the channels for consumers to shop online are also varied. The more accurate the search results provided by the technology team's machine learning system to consumers, the more likely consumers are to purchase. Because the information presented by search engines can largely guide these potential consumers. In the face of international consumer search requests, the addition of machine learning can make search results more effective because they are more likely to describe the product in its native language than the language of the place of origin. For example, if you are searching for a Burberry hand bag with a metal trim, the processing of Spanish and English will be different. eBay's translation capabilities have been incorporated into machine learning, which has alleviated the troubles of international consumers in non-English speaking countries. No matter where the goods are sold or what kind of language books are on the packaging, it is very important for consumers to understand the product description and understand their due value. According to Fain, eBay's search engine incorporates a "best match" algorithm that can analyze consumer's known information, which of the consumer's hottest search terms are, and what the consumer might buy. Fain said: "This is the best proof of eBay's large-scale application of machine learning, which is a powerful weapon for facilitating transactions." Take the “sewing machine†search example above, and the best match should be based on the price of the sewing machine. If the consumer is using a desktop, eBay can also display varying degrees of "best match" in the margins of the screen, along with the types of items that the consumer may search (antiques or collections). On smartphones or other mobile devices, the screen size is such that the presentation information is greatly limited. So the problem came, because according to Fain, 50% of eBay transactions are done by mobile devices. If consumers want to select other related categories, they can only click on the link to open a new page. In order to ensure the matching degree of search results, eBay has been working hard to make its results accurately reflect the market value of the products, and its search results are arranged in descending order of prices. Transaction credibility has always been a key topic for eBay. Machine learning can identify which indicators can reflect credibility (for example, the seller's trading volume) and which can't. Similarly, if the seller's bad reviews or other issues are more, the ranking of its stores in the search results will be automatically lowered by the system. eBay joins machine learning on its website to keep sellers happy and buyers rest assured. Fain said that every time he completes a transaction, he will ask: "Whether this transaction satisfies our criteria? Although these search results are all derived from the customer's relevant information, they are also determined by machine learning." Click on the massive information behind Fain remembered the five years he worked at Yahoo, when his main job was to handle web page traffic, and it was these web pages that could provide eBay users with rich information. He said that for the system, a single click of the year only means that consumers are interested in it, there is no more in-depth information. And now eBay's system can analyze the consumer's click stream, thus prompting consumers to consume. For machine learning systems, "ordering is a very powerful piece of information evidence," and eBay has been collecting consumer purchase orders to make it work. Currently, eBay Engineering Department also tries to use a machine learning system to identify the quality of pictures uploaded by merchants, and tells merchants what kinds of pictures are likely to guide consumer spending. It analyzes the consumer's perception of the picture through the parallel processing of powerful graphic picture processing units. Fain said: "Based on eBay's huge database excavated from real life, the system can guess more possibilities under the graphic picture processing unit's blessing." Although Fain did not disclose the number of servers used for machine learning, he stated that it was equivalent to a server farm. eBay found the machine learning system to be a very valuable tool, so it increased its investment in machine learning hardware, hired more professional and technical personnel, and launched more machine learning projects. "Machine learning is a big investment for us," Fain said. Via informationweek Recommended reading: Buy buy buy! eBay again acquires Israeli data analysis company SalesPredict eBay Acquires AI Big Data Analytics Company to Optimize e-commerce Search and Customer Experience Easy Electronic Technology Co.,Ltd , https://www.yxpcelectronicgroups.com