Invented by Jennifer A. Jacobi, Eric A. Benson, Gregory D. Linden, Amazon Technologies Inc

The market for the use of electronic shopping carts for personal recommendations has seen significant growth in recent years. As consumers increasingly turn to online shopping, the need for personalized recommendations has become crucial for businesses to stay competitive. Electronic shopping carts equipped with recommendation engines have emerged as a powerful tool to enhance the shopping experience and drive sales. Personalized recommendations have become an integral part of the online shopping journey. With the vast amount of products available, consumers often feel overwhelmed and struggle to find what they are looking for. This is where electronic shopping carts come into play. By analyzing user data, such as browsing history, purchase behavior, and demographic information, recommendation engines can suggest relevant products tailored to each individual’s preferences and needs. One of the key advantages of using electronic shopping carts for personal recommendations is the ability to offer a personalized shopping experience. By understanding the customer’s preferences and previous interactions, businesses can provide targeted suggestions that align with their interests. This not only saves time for the consumer but also increases the likelihood of making a purchase. Studies have shown that personalized recommendations can significantly boost conversion rates and customer satisfaction. Furthermore, electronic shopping carts with recommendation engines can also help businesses increase their average order value. By suggesting complementary or higher-priced items, businesses can encourage customers to add more products to their carts. This upselling technique has proven to be effective in increasing revenue and maximizing the value of each transaction. In addition to personalized recommendations, electronic shopping carts can also offer social recommendations. By integrating social media platforms, businesses can leverage user-generated content and recommendations from friends and influencers. This social aspect adds a layer of trust and authenticity to the recommendations, making them even more compelling for consumers. The market for electronic shopping carts for personal recommendations is not limited to any specific industry. From e-commerce giants to small businesses, companies across various sectors are adopting this technology to enhance their online shopping experience. Whether it’s fashion, electronics, or home decor, personalized recommendations can be applied to virtually any product category. However, it is important for businesses to handle user data responsibly and ensure privacy and security. With the increasing concern over data breaches and privacy issues, consumers are becoming more cautious about sharing their personal information. Therefore, businesses must be transparent about their data collection practices and provide clear opt-in options for users. In conclusion, the market for the use of electronic shopping carts for personal recommendations is thriving. As consumers seek personalized experiences and businesses strive to increase sales, recommendation engines have become an essential tool. By leveraging user data and offering tailored suggestions, businesses can enhance the shopping journey, increase customer satisfaction, and drive revenue. However, it is crucial for businesses to prioritize privacy and security to maintain consumer trust in this rapidly growing market.

The Amazon Technologies Inc invention works as follows

A computer-implemented services recommends products or items to a customer based on items that are known to be of interest, such as items in the customer’s electronic shopping basket. In one embodiment, a service identifies the items currently in the shopping cart of a user and uses them to generate a listing of additional items predicted to be interesting to the user. An additional item will then be selected to appear on the list depending in part on whether it is related to multiple items in the shopping cart. The item relationships are determined preferably by an offline process that analyzes the user’s purchase history to identify correlations. When the user is viewing the contents of their shopping cart, the additional items will be displayed.

Background for Use of electronic shopping carts for personal recommendations

A recommendation service” is an computer-implemented system that recommends products from a database. These recommendations are tailored to the user based on their information. Online customers are often recommended products by recommendation services. Online merchants offer services to recommend products (such as books, compact discs and videos) for example. Customers are recommended products based on the profiles created for them. It is also common to use recommendation services for users, who can be recommended Web sites, articles and other informational content.

Content-based filtering is a technique that recommendation services use. Pure content-based systems work by trying to identify items that, based upon an analysis of the item content, are similar items to those items known to be of user interest. A content-based Web page recommendation service, for example, may parse the user’s favourite Web pages in order to create a profile of frequently-occurring words, and then search other Web pages which include any or all of those terms.

Content-based system have significant limitations.” Content-based methods, for example, do not usually provide a mechanism to evaluate the popularity or quality of an item. Content-based methods also require that items have some kind of content that can be parsed by feature extraction algorithms. As a result, they are not suitable for movies, music, authors, restaurants and other items with little or no useful content.

Collaborative filtering is another common recommendation technique. In a pure collaboration system, users are given recommendations based on a group of users’ interests, but without any analysis of the item content. In collaborative systems, users are asked to rate items on a list. Each user creates a profile of their ratings through this process. In order to generate recommendations for one user, the profile of that user is compared with the profiles other users in order to identify a number of’similar users.’ The user is then shown items that have been highly rated by similar users, but which the user has not yet rated. The collaborative filtering method is a great way to overcome the shortcomings of content-based filters.

As with content-based filters, collaborative filtering has several issues. The user has to often rate items from the database in order to create a rating profile. This can be a frustrating task, especially if you are unfamiliar with the items being rated. Moreover, since collaborative filtering is dependent on other users with similar tastes, collaborative systems are not well suited to provide recommendations to users who have unusual taste.

Another issue with collaborative filtering is that an item from the database cannot be recommended until it has been rated. The operator of a collaborative recommendation system often faces a “cold start” as a result. The service can’t be launched in an effective form until enough ratings have been collected. Even after the service is brought online, it can take months or even years to recommend a significant number of items in the database.

Another problem with collaborative filtering methods is that the task of comparing user profiles tends to be time consuming?particularly if the number of users is large (e.g., tens or hundreds of thousands). In order to achieve a balance between speed and depth of analysis, there is a tendency for a trade-off. In a recommendation system, for example, that provides real-time suggestions in response to user requests, it might not be possible to compare a user’s rating profile with those of other users. It is therefore possible to perform a relatively superficial analysis of the data available (leading, in turn, to poor recommendations).

Another issue with collaborative and content-based system is that they do not generally reflect the current preferences within the community. There is no way to favor items that are ‘hot sellers’ in a system which recommends products. Existing systems also do not have a way to recognize that a user is searching for an item of a certain type or category.

The present invention provides a computer-implemented system and methods that generate personalized recommendations based on a group of users’ collective interests. The service has the important benefit that recommendations can be generated without the user or other users having to rate the items. The recommended items can be identified by using a table or mapping structure that has been previously generated. This maps each item to a list of “similar” items. items. The table reflects item similarity based on at least correlations of user interests in specific items.

The service can recommend a wide range of products, including books, compact disks (?CDs?) and videos. Videos, authors, musicians, item categories and Web sites are all possible. The service can be implemented as part of an online services network or e-mail notification system or any other computer system that recommends items explicitly or implicitly to users. In the preferred embodiment described in this document, the service is used by an online merchant to recommend titles such as music and book titles to its Web site users.

According to one aspect of the invention the mappings (?item-to item mappings?) of items are generated periodically, such as once per week, by an off-line process which identifies correlations between known interests of users in particular items. The mappings of items to similar items (?item-to-item mappings?) are generated regularly, for example, once a week, using an offline process that identifies correlations among known user interests in specific products. In the embodiments described below, for example, mappings are generated by periodically analyzing purchase histories of users to identify correlations. The number of users who have an interest both in the two items and the number that are interested in one item is preferred to measure the similarity. (e.g. items A and B have a high degree of similarity because a large number of those users bought each item). These mappings can also include other types of similarity, such as content-based similarities that are extracted from item descriptions or contents.

To create a set or recommendations for an individual user, the service will retrieve from the table similar items lists that correspond to items known to be relevant to the user and combine these lists appropriately to produce a list. If, for example, three items are known to interest the user (such a three items they recently purchased), then the service can retrieve and combine the lists of similar items. The item-to -item mappings, which are updated periodically using the most recent sales data available, tend to reflect current purchasing trends.

According to another aspect of this invention, similar items lists from the table can be weighted appropriately (before being combined) according to the indications of the user?s affinity or current interest for the corresponding items. The similar items list of a book purchased within the past week could be given a higher weight than the list of similar items for a four-month-old book. The weighting of a list of similar items increases the chances that items from that list are included in the final recommendations presented to the user.

An important aspect of the service is that the relatively computation-intensive task of correlating item interests is performed off-line, and the results of this task (item-to-item mappings) stored in a mapping structure for subsequent look-up. The personal recommendations can be generated quickly and efficiently, such as in real time in response to the user’s request.

Another aspect of the invention is the use of current or recent contents in the user’s cart to provide inputs to the service of recommendations (or another type of service that generates suggestions based on a list of unary items). If the user has three items currently in their shopping cart, then these items can be used to generate recommendations. The recommendations will be displayed automatically, when the user looks at the contents of the shopping basket. The current or recent contents of the shopping cart are used to generate recommendations that are closely correlated with the short-term interest of the user, even if this is different from their general preferences. This method is more likely to produce books with similar or the same topics if, for example, the user has recently added several books in the shopping cart that are related to a certain topic.

A computer-implemented method for recommending items to the user is one aspect of the invention.” The method consists of identifying the items currently in a user’s cart, and then using that list to predict additional items the user will be interested in. An additional item that is included in the list is chosen based on whether it is similar to multiple items within the cart. The user is shown the list of additional items when viewing the contents of their shopping cart.

Another aspect is the method for recommending products to an end-user. The method includes generating a database that maps individual products into sets of related product, where the relatedness of products is determined at least partly based on an automated analysis based on user purchase histories. The method also comprises identifying multiple products in the shopping cart of an individual user. The data structure is accessed for each product to identify the corresponding set related products. This identifies a plurality sets of related items. The sets are used to select the related products to recommend to the users based on the fact that a product is related to multiple products in a user’s cart.

Another aspect of the invention is that it allows the user to have multiple shopping baskets on a single account, such as for family members. It also generates recommendations specific to each shopping cart. The user may be asked to choose a specific shopping cart or set of carts, and then recommendations are generated based on items purchased from the carts. This feature allows users to receive recommendations that are based on the purpose or role (e.g. work versus pleasure), of a specific shopping cart.

Two specific implementations are revealed, which both generate personal recommendations by using the same table. In the first implementation the recommendations are based upon the items the user has recently rated or bought. The second implementation is based on what the user has in their shopping cart.

The invention’s various features and methods will be described now in the context a recommendation service. This includes two specific implementations of this service. It is used to suggest book titles, music titles and video titles to users of The disclosed methods, as will be appreciated by those with knowledge of the art, can be used to recommend non-physical items. The disclosed methods may be used, for example, to recommend authors, musicians, categories or groups titles, websites, chat groups and groups, movies, television programs, downloadable content and restaurants.

Throughout the description, reference will be made to various implementation-specific details of the recommendation service, the Web site, and other recommendation services of the Web site. These details are included to illustrate the preferred embodiments of this invention and not to limit its scope. The appended claims define the scope of the invention.

Click here to view the patent on Google Patents.