FR3105472A3 - Request analysis system equipped with an identification unit for robotic requests and multiple requests - Google Patents

Abstract

The invention relates to a request analysis computer system (100) configured to analyze a set of requests received by a product supplier computer system (200), the product supplier system (200) being connected to the product system. analyzing requests (100) and being configured to deliver a set of product recommendations in response to a request, each request comprising a set of parameters associated with the request. The query analysis system (100) is configured to provide a set of information from at least a portion of the set of parameters for each query. The request analysis system (100) includes a request processing unit (150), the request processing unit (150) being configured to detect requests from bots or multi-requests. Figure for the abstract: Fig. 1

Description

Request analysis system equipped with an identification unit for robotic requests and multiple requests

The invention relates generally to analysis systems for providing relevant information on requests made by users to a computer system supplier of products.

In a product provider computer system, a user can submit a request from a client device (user device) to a server device via a communications network to obtain a product. The query includes a set of query parameters that can be mandatory or optional.

The query parameters can then be analyzed by query analysis computer systems, which provide a synthetic view of the information associated with the query parameters.

IT systems supplying products can respond to thousands of user requests per second, which means that request analysis systems are able to analyze a significant number of requests (several hundred million requests per day).

However, a not insignificant part of the requests made to the computer systems supplying products are not carried out by real users, but by robots. More generally, more than half of the world's Internet traffic is generated by bots, whether they are "benevolent" bots (eg traffic monitoring bots or search engine bots) or "bots". malicious ”(for example pirate bots or spam bots).

Identifying "live" robots is a documented problem in the computer industry. These techniques are based, among other things, on session information, namely an IP address or any identifier which makes it possible to know that several requests come from the same user. However, this identification cannot be carried out when the robots make requests from online search engines (for example online travel agencies or flight comparators in an example of an application to the travel industry), themselves connected computer systems that supply products.

The query analysis systems therefore do not have relevant information on the queries made by users, due to the activity of bots.

Also, when a user makes a query to an online search engine, the online search engine may perform several similar queries, but which differ from the user's initial query. A grouping of similar queries, generated by the online search engine, following a query actually made by a user, is called a “multi-query”.

The traffic generated by bots and multi-queries greatly increases the volume of data to be processed by query analysis systems. It is therefore necessary to clean the data to be analyzed.

There is therefore a need for a query analysis system and method capable of detecting data actually generated traffic by users of a product supplier system.

The invention improves the situation. To this end, it proposes a request analysis computer system configured to analyze a set of requests received by a product supplier computer system, the product supplier system being connected to the request analysis system and being configured to deliver a request. set of product recommendations in response to a query, each query including a set of parameters associated with the query, the query analysis system being configured to provide a set of information from at least part of the set of parameters for each query. Advantageously, the request analysis system comprises a request processing unit, the request processing unit being configured to detect requests sent by robots or multi-requests.

In one embodiment, the request processing unit is configured to determine requests linearly correlated in a multidimensional space representing, over a determined period of time, the evolution over time of at least one parameter selected from among the set of parameters. parameters associated with the request.

The query processing unit can be configured to filter linearly correlated queries.

According to one particular aspect, the request processing unit can be configured to detect requests grouped together in the same time window in a multidimensional space representing, over a determined period of time, the evolution over time of the value of at least one. parameter selected from the set of parameters associated with the request.

According to another aspect, the request processing unit can be configured to replace the requests grouped in the same time window with a single request representative of the requests grouped in the same time window.

Further provided is a computer implemented method of analyzing requests received by a product supplier computer system, the method comprising the steps of:
- provision, by the computer system supplier of products, of a set of product recommendations in response to a request, each request comprising a set of parameters associated with the request,
- supply, by a request analysis computer system connected to the product supplier computer system, of a set of information from at least part of the set of parameters for each request,
characterized in that the method comprises a step of detecting, by a request processing unit, requests sent by robots or multi-requests.

In one embodiment, the detection step may comprise determining the requests linearly correlated in a multidimensional space representing, over a determined period of time, the evolution over time of at least one parameter selected from the set of associated parameters. at the request.

The method may further include a step of filtering linearly correlated queries among the set of queries.

The detection step can comprise the detection of the requests grouped together in the same time window in a multidimensional space representing, over a determined period of time, the evolution over time of the value of at least one parameter selected from the set of parameters. associated with the query.

The method can include a step of replacing the requests grouped together in the same time window by a single request representative of the requests grouped in the same time window.

Other characteristics, details and advantages of the invention will emerge on reading the description made with reference to the accompanying drawings given by way of example and which represent, respectively:

FIG. 1, a schematic illustration of the request analysis system according to the invention;

FIG. 2 a first illustration in three-dimensional view of requests made by robots;

FIG. 3, a second illustration in three-dimensional view of requests made by robots;

Figure 4, a three-dimensional view illustration of multi-queries;

FIG. 5, an illustration of the architecture of a computer capable of implementing the invention.

Figure 1 schematically illustrates the query analysis system 100 according to the invention. The request analysis system 100 retrieves, for example periodically, all of the requests received by a product supplier system 200 over a determined period, for example daily. The product provider system 200 is configured to deliver a set of product recommendations in response to a query.

In an example of application to the travel industry, the product supplier server device may be a GDS (“Global Distribution System”) type system allowing users to reserve and purchase products. such as plane tickets, train tickets or even hotel nights, by having direct access to the data provided by the travel supplier (which can be for example, an airline in the case of a reservation of plane ticket).

Each request retrieved by the request analysis system 100 includes a set of parameters associated with the request. The parameters of the request can be for example the origin, the destination, the date of the outward journey and the date of the return journey. The date of the return trip can be entered manually by the user, or by default be set to a date corresponding to a certain number of days after the date of the outbound trip (for example a week). The user can change the return travel date before validating the request. The parameters associated with each request can be numeric values, for example dates, or they can be character strings, for example departure and arrival airports.

The request analysis system 100 may communicate with a client device 300. The request analysis system 100 provides the client device 300 with a set of relevant information from at least a portion of the set of values of. parameters of each request. In an example of application to the travel industry, the relevant information may be a representative report,

In an example of an application to the travel industry, the parameters of the request can be for example the origin, the destination, the date of the outward journey and the date of the return journey. The date of the return trip can be entered manually by the user, or by default be set to a certain number of days after the date of the outward journey (for example one week), which the user can modify before validating his request.

The query parameters entered by the users can be retrieved by computer systems for query analysis, in order to obtain synthetic information on the centers of interest of the users, for example the origins / destinations which are most frequently entered. , the average time between the date on which the request is made and the date of the outward journey (time linked to the advance purchase condition), the average time between the outward journey and the return trip. This information can be provided globally, or be targeted in relation to a predetermined market (users making a request in a specific country, for example), or else a connection identifier, specific to each customer of the product supplier system. The information restored at the level of the client device 300 can constitute decision aids, for example for commercial operations, or else for statistical purposes.

Advantageously, the request analysis system 100 includes a request processing unit 150. The request processing unit 150 is configured to filter the linearly correlated requests. The request processing unit 150 is configured to detect requests that have not actually been sent by users, such as requests from robots, or multi-requests, for example. It is also configured to clean data after detecting data that does not come from real users (or humans).

FIG. 2 schematically illustrates a three-dimensional space representing a distribution of requests as a function of time over a day, according to two parameters associated with the requests. The space represented is three-dimensional, but it could alternatively be one-dimensional, two-dimensional, or more generally be multidimensional. In the example shown in Figure 2, the first parameter is the number of days (ADV) between the day of the request and the requested departure date. The second parameter is the length of stay entered in the request, namely the time between the "outbound" journey and the "return" journey. The queries displayed in three-dimensional space are associated with the same origin and the same destination. Parameters can be fixed between each of the requests, such as for example the user account identifier associated with the product supplier system 200. Parameters, other than the duration (ADV) which precedes the date of the outward journey and the length of stay entered. in the request, can be taken into account.

A grouping operation by correlation of the different requests can be carried out by the request processing unit 150. Several grouping algorithms (or "clustering" according to the English terminology) can be considered. In particular, the "COPAC" algorithm (COrrelation Partition Clustering) seems adapted to the field of application considered, in terms of execution time, flexibility and quality of results. As a variant, other algorithms could be used, such as the ORCLUS, 4C, HiCO or ERiC algorithms, all of which are known to those skilled in the art. All of these methods use the principle of local principal component analysis ("Local PCA"). It is possible, with the aforementioned methods, to detect the requests which are linearly correlated with one another.

FIG. 2 represents traffic typically generated exclusively by robots. For example, the data series located in the upper part of the three-dimensional space represents a series of requests for the same origin, the same destination, and from the same user account of the product supplier system 200. At very short time intervals , and for an identical length of stay, the robot can make requests for which only the time between the departure date and the request date varies. The regular pattern in the form of a straight line of the series of queries corresponds to a linear correlation of the different queries in the series.

The system according to the invention is also capable of detecting mixed traffic, namely generated by real users, and by robots, from the same user account. For example, in figure 3, a series of points appearing in the box reveal a linearity corresponding to similar requests, brought closer in time, and varying one with respect to the other only by the parameter linked to length of stay.

The request processing unit 150 can be configured to filter the requests exhibiting a strong linear correlation between them by suppressing them, which makes it possible to significantly reduce the volume of data to be processed by the request analysis system 100. request processing unit 150 is thus configured to determine requests linearly correlated in a multidimensional space representing, over a determined period of time, the evolution over time of at least one parameter selected from the set of parameters associated with the request.

The query processing unit 150 can also be configured to detect multi-queries, i.e. similar queries, with search criteria that differ little from each other. The multi-query detection algorithm can include a time series analysis, in order to group together similar queries that have been performed in the same time window, for example a Gaussian window.

For multi-query detection, the data cleansing process does not remove all queries from a multi-query, because one query among all queries was actually made by a user. The data cleaning process is thus implemented to keep only one query from the set considered to be a multi-query. For this, the convolution product of the series of queries with a time window defined previously can be calculated. The query corresponding to a maximum of the convolution product is considered to be the query representing all the queries of the multi-query.

FIG. 3 illustrates a representation of the processing of multiple requests in a three-dimensional space. The three-dimensional space represents the same parameters as in Figures 2 and 3, namely the time of day when the request was made (in seconds), the time between the start date and the request (in days), and the length of stay indicated in the request. Here too, other parameters can be taken into consideration, without departing from the scope of the invention. Similar requests forming a multi-request are grouped together by processing unit 150. The left part of Figure 4 represents four multi-requests. The right part of Figure 4 shows the queries in three-dimensional space, after data cleaning. For each multi-query, a single so-called representative query was kept by processing unit 150.

The request processing unit 150 can thus be configured to detect requests grouped together in the same time window in a multidimensional space representing, over a determined period of time, the evolution over time of the value of at least one parameter selected from among the set of parameters associated with the request.

The request processing unit 150 can also be configured to replace the requests grouped together in the same time window with a single request representative of the requests grouped together in the same time window.

The volume of data to be processed by the query analysis system 100 is also significantly reduced, while maintaining an objective view of the reality of user interest in the various products offered by the product supplier system 200.

Furthermore, the data cleaning performed by the processing unit 150 guarantees a suitable response time of the request analysis system 100.

The aforementioned computer systems, processing units and client devices can be integrated into one or more computers as shown in Figure 5.

The computer can include a processor, a memory, a mass storage memory device, an input / output (I / O) interface and a Human Machine Interface (HMI). The computer can also be operably coupled to one or more external resources via the network and / or an I / O interface. External resources may include, but are not limited to, servers, databases, mass storage devices, peripheral devices, cloud-based network services, or any other suitable computing resource that can be used by the computer.

The processor may include one or more devices selected from microprocessors, microcontrollers, digital signal processors, microcomputers, central processing units, user-programmable pre-broadcast networks, programmable logic devices, digital processing machines. state, logic circuits, analog circuits, digital circuits, or other devices that manipulate signals (analog or digital) according to operating instructions that are stored in memory. The memory may include a single memory device or a plurality of memory devices including, but not limited to, read only memory (ROM), random access memory (RAM), volatile memory, nonvolatile memory, random access memory. static (SRAM), dynamic random access memory (DRAM), flash memory, cache memory, or any other device capable of storing information. The mass storage memory device may include data storage devices such as a hard drive, an optical disc, a cassette player, a non-volatile semiconductor device, or any other device capable of storing information. . A database can reside on the mass storage memory device, and can be used to collect and organize data used by the various systems and modules described herein. The processor can operate under the control of an operating system that resides in memory.

The operating system can manage computer resources so that embedded computer program code in the form of one or more software applications, such as a memory resident application, can have instructions executed by the processor. . In another embodiment, the processor can run the application directly, in which case the operating system can be omitted. One or more data structures may also reside in memory, and may be used by the processor, operating system, and / or application for data storage or manipulation.

The I / O interface can provide a machine interface that functionally couples the processor to other devices and systems, such as the network and / or an external resource. The application can thus work in cooperation with the network and / or an external resource by communicating via the I / O interface to provide the various features, functions, applications, processes and / or modules comprising embodiments of the invention. . The application may also have program code that is executed by one or more external resources, or otherwise rely on functions and / or signals provided by other system or network components external to the computer. Indeed, given the almost infinite number of possible hardware and software configurations, those skilled in the art will understand that the embodiments of the invention may include applications which are located outside the computer, distributed over several computers. or other external resources, or provided by computing resources (hardware or software), in the form of a service on the network, such as a cloud computing service.

The HMI interface can be operably coupled to the processor of the computer in a known manner, to allow a user of the computer to interact directly with the computer. The HMI interface may include video and / or alphanumeric screens, a touch screen, a speaker, and any other appropriate audio and visual indicator capable of providing information to the user. The HMI interface can also include input devices and commands such as an alphanumeric keyboard, pointing device, numeric keypads, push buttons, control buttons, microphones, etc., capable of receiving commands. and input data from the user and transmit the inputs entered into the processor.

Claims (10)

Request analysis computer system (100) configured to analyze a set of requests received by a product supplier computer system (200), the product supplier system (200) being connected to the request analysis system (100) and being configured to deliver a set of product recommendations in response to a query, each query including a set of parameters associated with the query, the query analysis system (100) being configured to provide a set of information from 'at least a part of the set of parameters of each request, characterized in that the request analysis system (100) comprises a request processing unit (150), the request processing unit (150) being configured to detect requests emitted by robots or multi-requests. System according to Claim 1, characterized in that the request processing unit (150) is configured to determine linearly correlated requests in a multidimensional space representing, over a determined period, the evolution over time of at least one parameter selected from the set of parameters associated with the request. System according to claim 2, characterized in that the request processing unit (150) is configured to filter the linearly correlated requests. System according to Claim 1, characterized in that the request processing unit (150) is configured to detect requests grouped together in the same time window in a multidimensional space representing, over a determined period, the evolution over time of the value of at least one parameter selected from the set of parameters associated with the request. System according to Claim 4, characterized in that the request processing unit (150) is configured to replace the requests grouped together in the same time window by a single request representative of the requests grouped in the same time window. A computer-implemented method of analyzing requests received by a product supplier computer system (200), the method comprising the steps of
- provision, by the product supplier computer system (200), of a set of product recommendations in response to a request, each request comprising a set of parameters associated with the request,
- supply, by a request analysis computer system (100) connected to the product supplier computer system (200), of a set of information from at least part of the set of parameters of each request ,
characterized in that the method comprises a step of detecting, by a request processing unit (150), requests sent by robots or multi-requests. Method according to Claim 6, characterized in that the detection step comprises determining the requests linearly correlated in a multidimensional space representing, over a determined period of time, the evolution over time of at least one parameter selected from the set. parameters associated with the request. Method according to claim 7, characterized in that the method comprises a step of filtering linearly correlated requests among the set of requests. Method according to Claim 6, characterized in that the detection step comprises the detection of the requests grouped together in a same time window in a multidimensional space representing, over a determined period, the evolution over time of the value of at least a parameter selected from the set of parameters associated with the request. Method according to Claim 9, characterized in that it comprises a step of replacing the requests grouped together in the same time window by a single request representative of the requests grouped in the same time window.
………………….