JP7092695B2 - Property search support system and property search support method - Google Patents

Description

The present invention relates to a property search support system and a property search support method, and specifically relates to a technique that enables efficient and accurate real estate property search regardless of the skill and experience of the person in charge.

Real estate sales representatives search for a property that meets the desired conditions specified by the customer, and present the property as a contract candidate to the customer. Appropriate experience and skills are required to efficiently search for properties that meet the customer's desired conditions from a huge number and types of properties for sale.

Therefore, various assistive techniques related to such property search work have been proposed. For example, it is a real estate transaction support system that supports transactions between a person who wants to purchase real estate and a person who wants to sell it, and registers the purchase conditions of the person who wants to purchase, which is input from a communication terminal, in association with the contact information of the person who wants to purchase. The purchase condition registration means, the property information registration means for registering the property information provided by the purchase applicant in association with the contact information of the purchase applicant, and the purchase condition of the purchase applicant in the property information registration means are met. Specific property information that associates the property information with the purchase applicant at the time of matching, in which the property information determination means for determining whether or not the property information is registered and the property information that matches the purchase conditions of the purchase applicant are registered. A real estate transaction support system (see Patent Document 1) equipped with a specific property information storage means for storing specific property information and a contact means for contacting the contact information of a person who wishes to purchase the specific property information stored in the specific property information storage means. Proposed.

In addition, a calculation unit that calculates the feature amount of the real estate property or the event related to the real estate property, the closing period in the past successful transaction, the feature amount of the target real estate property, and the feature amount of the current target real estate property of the transaction. Based on the above, an information processing device (see Patent Document 2) including a prediction unit for predicting a contract probability of a predetermined transaction period in the transaction has also been proposed.

JP-A-2017-117248 Japanese Unexamined Patent Publication No. 2017-16321

On the other hand, the desired conditions specified by the customer are often ambiguous without specific numerical values, or can be interpreted in multiple ways by expressions in natural language.

In such a case, even if the above-mentioned persons in charge perform a property search in a predetermined DB, it is difficult to set the search conditions themselves, and naturally the obtained search results tend to be difficult to obtain the desired accuracy.

Then, the business efficiency related to the property search tends to decrease more and more, and for those who lack experience and skills, it is not possible to secure the target property of the customer proposal, and it becomes difficult to carry out the business.

On the other hand, even if an experienced person in charge searches for a property, if the result is that there are no properties for sale that meet the customer's desired conditions, the property is searched while personally considering conditions that are partially similar to the desired conditions. It is necessary to do this, and it requires a certain degree of difficulty and further effort as a business.

Therefore, an object of the present invention is to provide a technique that enables efficient and accurate search for real estate properties regardless of the skill and experience of the person in charge.

The property search support system of the present invention that solves the above problems learns a storage device that holds selling information, buying information, and contract information of a real estate property, and a pair of selling information and buying information indicated by the contract information. Processing as data, processing to specify the range of difference between the desired condition indicated by the buying information in the contract information and the condition of the actual contracted real estate property, buying information and selling information in which the desired condition is included in the difference range and the sale is unsuccessful. A computing device that executes a process of using the pair as negative example learning data, and a process of executing machine learning based on the positive example learning data and the negative example learning data to generate a contract probability model for buying and selling real estate. include.

Further, in the property search support method of the present invention, the information processing system holds the selling information, the buying information, and the contract information of the real estate property in the storage device, and the pair of the selling information and the buying information indicated by the contract information is obtained. Processing as regular learning data, processing to specify the range of difference between the desired condition indicated by the buying information in the contract information and the condition of the actual contracted real estate property, buying information in which the desired condition is included in the difference range and the sale is unsuccessful And a process of using a pair of selling information as negative example learning data, and a process of executing machine learning based on the positive example learning data and the negative example learning data to generate a contract probability model for buying and selling real estate.

According to the present invention, it is possible to efficiently search for real estate properties with good accuracy regardless of the skill and experience of the person in charge.

It is a network configuration diagram including the property search support system of this embodiment. It is a figure which shows the hardware configuration example of the property search support system in this embodiment. It is a figure which shows the data structure example of the selling information DB of this embodiment. It is a figure which shows the data structure example of the buying information DB of this embodiment. It is a figure which shows the data structure example of the contract information DB of this embodiment. It is a figure which shows the structural example of the contract probability model of this embodiment. It is a figure which shows the data structure example of the pattern dictionary of this embodiment. It is a figure which shows the data structure example of the contradiction resolution rule of this embodiment. It is a figure which shows the data structure example of the conversion dictionary of this embodiment. It is a figure which shows the flow example 1 of the property search support method in this embodiment. It is a figure which shows the screen example 1 in this embodiment. It is a figure which shows the screen example 2 in this embodiment. It is a figure which shows the flow example 2 of the property search support method in this embodiment. It is a figure which shows the conceptual example of the regular example learning data in this embodiment. It is a figure which shows the conceptual example of the negative example learning data in this embodiment. It is a figure which shows the conceptual example of the contract probability model target in this embodiment. It is a figure which shows the screen example 2 in this embodiment.

－－－ Network configuration －－－
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a network configuration diagram including the property search support system 100 of the present embodiment. The property search support system 100 shown in FIG. 1 is a computer system that enables efficient and accurate search for real estate properties regardless of the skill and experience of the person in charge.

Such a property search support system 100 can assume, for example, a server device operated by an organization such as a real estate company or a financial institution that sells and sells real estate properties and conducts brokerage business. However, the operation mode of the property search support system 100 is not limited and can be adopted in various ways.

The above-mentioned property search support system 100 is connected to a user terminal 200 operated by a sales representative of the above-mentioned real estate company, for example, via an appropriate network 10 such as the Internet or an intranet so as to be capable of data communication.

The user terminal 200 can be assumed to be a mobile terminal such as a smartphone or a tablet terminal in addition to a general PC. The user terminal 200 holds buy information 225 and sell information 226 heard from the customer by the sales person.

The above-mentioned purchase information 225 is information that specifies the conditions of the real estate property that the customer wants to purchase. Further, the sale information 226 is information that describes the conditions of the real estate property that the customer wants to sell. In either case of buy information 225 or sell information 226, specifically, for example, the location, the nearest station, the distance from the nearest station, the land area, the presence or absence of a building, the total floor area of the building, and the road contact status. , Etc. are included.

The sales person operates the user terminal 200 and transmits the purchase information 2254 and the sale information 226 heard from the customer to the property search support system 100 via the network 10. Therefore, the property search support system 100 of the present embodiment acquires the buy information 225 and the sell information 226 in the user terminal 200 according to the operation of the above-mentioned sales person or the like or at predetermined intervals, and obtains the own storage device 101. It is assumed that the buying information DB 111 and the selling information 110 in the above are stored and managed.

--- Hardware configuration ---
Further, the hardware configuration of the property search support system 100 of the present embodiment is as shown in FIG. That is, the property search support system 100 includes a storage device 101, a memory 103, an arithmetic unit 104, an input device 105, an output device 106, and a communication device 107.

Of these, the storage device 101 is composed of an appropriate non-volatile storage element such as an SSD (Solid State Drive) or a hard disk drive.

Further, the memory 103 is composed of a volatile storage element such as a RAM.

Further, the arithmetic unit 104 is a CPU that executes the program 102 held in the storage device 101 by reading it into the memory 103, etc., to perform integrated control of the apparatus itself, and to perform various determinations, arithmetic operations, and control processes.

Further, the input device 105 is a device such as a keyboard, a mouse, and a microphone that accepts key input and voice input from the user.

Further, the output device 106 is a device such as a display that displays processed data.

Further, the communication device 107 can be assumed to be a network interface card that is connected to an appropriate network 10 such as the Internet and is responsible for communication processing with other devices such as the user terminal 200.

In addition, in the program 102 for implementing the function required as the property search support system of the present embodiment, at least the machine learning engine 1021 is stored in the storage device 101 in addition to the general device control program. ing.

This machine learning engine 1021 is an engine that executes machine learning by inputting positive example learning data and negative example learning data generated based on the contract information DB 112 and generates a contract probability model for real estate sales. It should be noted that the regular learning data is generated from the pair of the selling information and the buying information indicated by the contract information in the contract information DB 112. Further, in the negative example learning data, the desired condition is included in the difference range between the desired condition indicated by the buying information in the contract information DB 112 and the condition of the actual contracted real estate property, and the buying information and the selling information of the unsuccessful sale are included. Generated from a pair.

Further, in addition to the above-mentioned program 102, at least the selling information DB 110, the buying information DB 111, the contract information DB 112, the contract probability model 113, the pattern dictionary 114, the contradiction resolution rule 115, and the conversion dictionary 116 are stored in the storage device 101. Has been done. However, the details of these various types of information will be described later.

--- Data structure example ---
Subsequently, the tables used by the property search support system 100 of the present embodiment will be described. FIG. 3 shows an example of the selling information DB 110 in the present embodiment.

The sale information DB 110 is a database that stores the sale information handled / handled by the real estate company or the like supported by the property search support system 100.

The data structure is a collection of records consisting of data related to various items such as property type, seller, location, nearest station, land area, building, total floor area, contract status, etc., using the property ID as a key. Of course, each item illustrated here is a part, and it is natural that other items required for real estate sales are also included. This point is the same for the items in the purchase information DB 111.

Further, FIG. 4 shows an example of data configuration of the buying information DB 111 of the present embodiment. The purchase information DB 111 of the present embodiment is a database that stores the purchase information handled / handled by the real estate company or the like supported by the property search support system 100.

The data structure consists of data such as the customer ID of the customer who wants to purchase the property, the desired property location, the nearest station, the land area, the building, the total floor area, the contract status, and the remarks, using the property ID as a key. It is a collection of records.

Of these, the one whose contract status is "not yet" is related to the purchase information received from the customer by the sales person, and is the purchase information to be processed by the property search support system 100 of the present embodiment.

In addition, the remarks are text data described in natural language with detailed desired contents specifically stated by the customer regarding the purchase information.

Further, FIG. 5 shows an example of data configuration of the contract information DB 112 of the present embodiment. The contract information DB 112 of the present embodiment is a database that stores information on real estate for which a sales contract has been concluded by matching with the purchase information among the selling information.

The data structure uses the contract ID as a key, and corresponds to the sale information and the purchase information that led to the contract, such as property ID, property type, location, nearest station, land area, building, total floor area, and remarks. A collection of records consisting of data. For the selling information in which the information is stored in the contract information DB 112, the value of "completed" is set in the "contract status" column in the selling information DB 110. Further, the buying information whose information is stored in the contract information DB 112 is
In the purchase information DB 111, the value of "completed" is set in the "contract status" column.

Further, FIG. 6 shows a configuration example of the contract probability model 113 of the present embodiment. The contract probability model 113 of the present embodiment is a database accumulating the contract probability model generated by the property search support system 100.

The data structure is a collection of records consisting of data such as a contract probability calculation target and a decision tree file, using a model ID that uniquely identifies the contract probability model as a key. Here, an example in which a decision tree is adopted as an implementation form of the contract probability model is shown, but of course this is an example, and another algorithm or the like capable of calculating the contract probability may be appropriately adopted.

Further, FIG. 7 shows an example of the data structure of the pattern dictionary 114 of the present embodiment. The pattern dictionary 114 of the present embodiment is a database accumulating normalization rules for descriptions in natural language.

The data structure is a collection of records composed of data such as extraction rules and meanings, using a pattern ID that uniquely identifies the normalization rule as a key.

Of these, the extraction rule defines the text to which this pattern dictionary is applied, that is, the vocabulary included in the value in the remarks column of the buying information, the combination of vocabularies, or the phrase, which can be recognized ambiguously. There is. As this extraction rule, for example, it can be assumed that a person in charge with knowledge refers to various buying information in advance and sets it based on the description.

Further, the "meaning" is defined as a value that uniquely specifies the meaning of the vocabulary, phrase, etc. defined in the above-mentioned extraction rule column. As for the value in the "meaning" column, it can be assumed that the person in charge with the above-mentioned knowledge refers to various buying information in advance and sets the value based on the description.

Further, FIG. 8 shows an example of data configuration of the contradiction resolution rule 115 of the present embodiment. The contradiction resolution rule 115 of the present embodiment deals with a case where a contradiction occurs between the "meanings" uniquely specified in the above-mentioned pattern dictionary regarding the value in the remarks column of the buying information, and gives priority to any of the meanings. It is a database that accumulates rules for deciding.

The data structure is a collection of records associated with meanings that should be preferentially adopted for the pair, that is, values of priority meanings, with meanings 1 and 2, which are pairs of meanings to be resolved inconsistencies, as keys. ..

Further, FIG. 9 shows an example of data structure of the conversion dictionary 116 of the present embodiment. The conversion dictionary 116 of the present embodiment is a database accumulating conversion rules for search conditions described in natural language.

The data structure is a collection of records associated with a search condition based on the intention of the meaning, using the "meaning" of the conversion target (the one specified by the pattern dictionary 114 and the contradiction resolution rule 115) as a key. In the search condition column, it is stipulated which of the search conditions of "search for land" and "search for land and building" should be "priority target" or "target" or "excluded". There is. Of course, this form is just an example, and the granularity of information such as "search for office building", "search for parking lot", "search for factory site", etc., according to the type of real estate property to be handled and the needs of customers, etc. It may be changed as appropriate.

－－－ Flow example 1 －－－
Hereinafter, the actual procedure of the property search support method in the present embodiment will be described with reference to the figure.
Various operations corresponding to the property search support method described below are realized by a program read into a memory or the like by the property search support system 100 and executed. Then, this program is composed of the code for performing various operations described below.

FIG. 10 is a diagram showing a flow example 1 of the property search support method in the present embodiment. Here, a process of determining an appropriate search condition based on the purchase information newly presented by the customer to the person in charge or the like will be described.

In this case, the above-mentioned person in charge operates the user terminal 200 and inputs the buying information indicated by the customer. On the other hand, the user terminal 200 accepts the input of the purchase information by the person in charge on the predetermined screen (screen 1100 in FIG. 11), and transmits the determination request of the search condition including the purchase information 225 to the property search support system 100 ( s1).

On the other hand, the property search support system 100 receives the purchase information from the above-mentioned user terminal 200, executes morphological analysis on the value in the remarks column of the purchase information, and decomposes it into vocabulary and phrases (s2). For example, if the value in the remarks column of the purchase information is the text "I want a property with an existing building, but I can consider only the land", "existing building", "yes", "property", "desired". , "Land only", "Examination", "OK".

Subsequently, the property search support system 100 applies the vocabulary or the like obtained in s2 described above to the pattern dictionary 114 to normalize it, and extracts the meaning included in the description (s3). When the value obtained in s2 is applied to the pattern dictionary 114 of FIG. 7, the meaning “land is also examined” is extracted by the extraction rule of the pattern ID “E0001”. Further, the meaning "even if there is an existing building" is extracted by the extraction rule of the pattern ID "E0005".

Further, when the property search support system 100 has a plurality of meanings extracted in s3, each pair that can be combined is applied to the contradiction resolution rule 115, and the meaning of one of the pairs is narrowed down (s4).

If the contradiction resolution rule 115 is applied to the pair of "existing building" and "land is also considered" in the above example, it is possible to narrow down the pair to give priority to "land is also considered".

Subsequently, the property search support system 100 applies the meaning unique to the narrowing down of s4 to the conversion dictionary 116 to determine the search condition, and returns this to the above-mentioned user terminal 200 (s5). For example, when the meaning "Land is also examined" obtained by narrowing down s4 is applied to the conversion dictionary 116 of FIG. 9, the search conditions are "target" for land and "priority target" for land and buildings. It will be decided.

On the other hand, the user terminal 200 displays the information of the search condition transmitted from the property search support system 100 (s6), and ends the process. In this case, FIG. 12 shows an example of the screen 1200 displayed on the user terminal 200.

According to the above flow, even if the desired condition specified by the customer can be interpreted in a plurality of ways by the expression in natural language, it is possible to efficiently convert it into an appropriate search condition. As a result, it is possible to search for real estate properties more efficiently and with high accuracy regardless of the skill and experience of the person in charge.

－－－ Flow example 2 －－－
Subsequently, although the property search in the sales information DB 110 is executed using the search condition determined in the above flow example 1, even if the real estate property that matches the search condition is not hit, it is expected to be retroactive to the customer. The process of identifying an appropriate real estate property that can be done will be described.

FIG. 13 is a diagram showing a flow example 2 of the property search support method in the present embodiment. In this case, first, the property search support system 100 uses the pair of the selling information and the buying information indicated by the contracted item in the contracted information DB 112 as regular learning data, and stores it in, for example, the memory 103 (s10).

This positive example learning data is learning data that is input to the machine learning engine 1021 together with the negative example learning data described later. Further, the concept of regular learning data is illustrated in FIG. 14A. This concept is positive, for example, when plotting conditions in sell information and buy information (eg, land area, total floor area, etc.), where the horizontal axis is the value in the sell information and the vertical axis is the value in the buy information. Example It is shown that the range where the training data is distributed is the gray spindle-shaped region. If the values of the corresponding conditions completely match the sell information and the buy information, or if the values are close to each other, it is presumed that the values converge to such a spindle-shaped region.

Subsequently, the property search support system 100 specifies a range of difference between the desired condition indicated by the buying information of the contracted item in the contract information DB 112 and the condition indicated by the selling information, that is, the condition of the actual contracted real estate property (s11).

For example, in the case of the constraint project "S0001" in the contract information DB 112 of FIG. 5, the difference range is between "land and building" and "land" with respect to the property type, and "C city D town" and "C" with respect to the location. Between "City F Town", between "W Line D Station" and "W Line E Station" for the nearest station, between "820" and "850 ~" for the land area, and "Yes" and "Yes" for the building. Each of "undesignated" and "1940" and "undesignated" with respect to the total floor area is specified as a difference range.

Subsequently, in the property search support system 100, the desired condition is included in the difference range specified in s11 described above, and the purchase information of the unsuccessful sale is obtained from the purchase information DB 110. Similarly, the desired condition is included in the difference range and the sale is unsuccessful. Selling information is extracted from the selling information DB 111, and pairs of buying information and selling information are generated, for example, in a predetermined number or as much as possible, and stored in the memory 103 as negative learning data (s12).

It can be expected that the buying information and the selling information, which include the condition in this difference range and the trading is unsuccessful, will be the trading establishment according to the correct example of the contract information DB 112, but in reality, the trading is unsuccessful. It is a thing.

The concept of negative example learning data is illustrated in FIG. 14B. This concept is negative, for example, when plotting conditions in sell information and buy information (eg, land area, total floor area, etc.), the horizontal axis is the value in the sell information and the vertical axis is the value in the buy information. Example It shows that the area where the training data is distributed is a gray rectangular area. If the value of the corresponding condition is separated between the sell information and the buy information and the sale is unsuccessful, it is presumed that the value converges to such a rectangular area. However, since it is limited to the range of difference, it does not disperse in a disorderly manner and is a region limited to a certain size.

Further, the property search support system 100 reads the positive example learning data obtained in s10 and the negative example learning data obtained in s12 from the memory 103, respectively, and inputs these to the machine learning engine 1021 for machine learning. Is executed to generate a contract probability model for buying and selling real estate (s13).

The generation of this contract probability model compares, for example, the conditions of selling information and buying information in the positive learning data (eg, land area, total floor area, etc.) with the conditions of selling information and buying information in the negative learning data. By doing so, the range of the difference range that truly contributes to the conclusion of the sale, that is, the range of the value of the sell information and the value of the buy information (see FIG. 14C) to be a positive example is specified, and the range is the range of the value of the condition. It is possible to envision a method of sequentially specifying the decision tree specified in the above using an existing generation algorithm. Of course, by comparing the conditions of selling information and buying information in the positive learning data (eg land area, total floor area, etc.) with the conditions of selling information and buying information in the negative learning data, it is certainly a positive example. In the learning data (that is, the contracted property that was closed for sale), the condition is that the selling information and the buying information always match, but the tendency is common to the negative learning data, and it is concluded that it is not an important item for closing the sale. Processing such as identifying conditions that are less important for determining the probability is also performed.

If a plurality of populations of the above-mentioned positive example learning data and negative example learning data are generated, a plurality of contract probability models are also generated for each population. In that case, the property search support system 100 generates one contract probability model by performing predetermined statistical processing on the plurality of contract probability models generated in s13. As this statistical processing, it is assumed that, for example, a majority vote, an average value, and a median value are taken.

Subsequently, in the property search support system 100, the conditions of the purchase information received from the customer (specified up to s5 in Flow Example 1) and the contract status in the sale information DB 110 are checked with respect to the contract probability model obtained in s13. The contract probability of the selling information is determined by applying the conditions of each selling information "not yet" (s14).

Further, the property search support system 100 outputs the information of the contract probability determined in s14 to the above-mentioned user terminal 200 as a customer proposal list (see screen 1500 in FIG. 15) (s15).

On the other hand, the user terminal 200 displays the customer proposal list delivered from the property search support system 100 (s16), and ends the process. The person in charge or the like who browses this customer proposal list can present the sales information to the customer, which can specify the appropriate contract probability, even if there is no sales information that exactly matches the desired conditions instructed by the customer.

Although the best mode for carrying out the present invention has been specifically described above, the present invention is not limited to this, and various modifications can be made without departing from the gist thereof.

According to this embodiment, it is possible to search for real estate properties efficiently and accurately regardless of the skill and experience of the person in charge.

The description herein reveals at least the following: That is, in the property search support system of the present embodiment, the arithmetic unit generates a plurality of the contract probability models and generates one contract probability model by performing predetermined statistical processing on the plurality of contract probability models. May be.

According to this, a contract probability model is appropriately obtained by performing statistical processing for obtaining an average or a median among a plurality of types of contract probability models obtained from a combination of various positive and negative training data. It is possible to narrow down to. As a result, it is possible to search for real estate properties more efficiently and with high accuracy regardless of the skill and experience of the person in charge.

Further, in the property search support system of the present embodiment, the arithmetic device sets the conditions of the buy information received from the customer and the conditions of the unsuccessful sale information held in the storage device with respect to the contract probability model. It may be applied to determine the closing probability of the selling information and further execute a process of outputting the closing probability information to a predetermined device.

According to this, for example, it is possible to generate a property list for customer proposals extracted in order from the one with the highest probability of contracting among the selling information currently held, and by extension, regardless of the skill and experience of the person in charge. It will be possible to search for real estate properties more efficiently and with good accuracy.

Further, in the property search support system of the present embodiment, the storage device has a pattern dictionary that defines a normalization rule for descriptions in natural language, a contradiction resolution rule that defines a unique rule for contradiction expressions in descriptions in natural language, and a contradiction resolution rule. Further holding a conversion dictionary that defines conversion rules for the description in natural language to the search condition, the arithmetic unit applies the description regarding the condition of the purchase information received from the customer to the pattern dictionary and normalizes the description. Obtained by the process of extracting the meaning to be included, the process of applying the contradiction resolution rule to the pair whose meanings are inconsistent among the extracted meanings, and the process of narrowing down to one of the meanings of the pair. The process of applying the meaning and the meaning unique by the narrowing down to the conversion dictionary and converting it into a search condition is further executed, and the search condition obtained by the conversion is used as the condition of the purchase information, and the condition of the purchase information is used. And the condition of the unsuccessful sale information held in the storage device is applied to the contract probability model, the contract probability of the sale information is determined, and the information of the contract probability is output to the predetermined device. May be further executed.

According to this, even if the desired condition specified by the customer can be interpreted in a plurality of ways by the expression in natural language, it is possible to efficiently convert it into an appropriate search condition. As a result, it is possible to search for real estate properties more efficiently and with high accuracy regardless of the skill and experience of the person in charge.

Further, in the property search support method of the present embodiment, the information processing system generates a plurality of the contract probability models and performs predetermined statistical processing on the plurality of contract probability models to generate one contract probability model. , May be.

Further, in the property search support method of the present embodiment, the information processing system determines the condition of the purchase information received from the customer and the condition of the unsuccessful sale information held in the storage device with respect to the contract probability model. May be applied to determine the contract probability of the sale information and further execute the process of outputting the contract probability information to a predetermined device.

Further, in the property search support method of the present embodiment, the information processing system defines a pattern dictionary that defines normalization rules for descriptions in natural language and unique rules for inconsistent expressions in descriptions in natural language in the storage device. It further holds a conversion dictionary that defines the inconsistency resolution rule and the conversion rule to the search condition of the description in natural language, and applies the description about the condition of the buying information received from the customer to the pattern dictionary to normalize the description. The process of extracting the meaning included in the information, the process of applying the contradiction resolution rule to the pair of the extracted meanings whose meanings are inconsistent with each other, and the process of narrowing down to one of the meanings of the pair, obtained by the extraction. The process of applying the meanings and the meanings unique to the narrowing down to the conversion dictionary and converting them into search conditions is further executed, and the search conditions obtained by the conversion are used as the conditions of the purchase information, and the purchase information of the purchase information is used. The condition and the condition of the unsuccessful sale information held in the storage device are applied to the contract probability model to determine the contract probability of the sale information, and the information of the contract probability is output to the predetermined device. Further processing may be performed.

10 Network 100 Property search support system 101 Storage device 102 Program 1021 Machine learning engine 103 Memory 104 Arithmetic device 105 Input device 106 Output device 107 Communication device 110 Sales information DB
111 Buying information DB
112 Contract information DB
113 Contract probability model 114 Pattern dictionary 115 Contradiction resolution rule 116 Conversion dictionary 200 User terminal

Claims (8)

A storage device that holds real estate property sale information, buy information, and contract information,
Processing that uses the pair of selling information and buying information indicated by the contract information as regular learning data, processing that specifies the range of difference between the desired condition indicated by the buying information in the contract information and the condition of the actual contracted real estate property, the difference A process of using a pair of buy information and sell information that includes desired conditions in the range and is unsuccessful in buying and selling as negative learning data, and machine learning based on the positive learning data and the negative learning data are executed to buy and sell real estate. An arithmetic unit that executes the process of generating a contract probability model, and
Property search support system including. The arithmetic unit is
A plurality of contract probability models are generated, and one contract probability model is generated by performing predetermined statistical processing on the plurality of contract probability models.
The property search support system according to claim 1, characterized in that. The arithmetic unit is
The contract probability of the sold information is determined by applying the condition of the buy information received from the customer and the condition of the unsuccessful sale information held in the storage device to the contract probability model, and the contract probability is determined. Further executes the process of outputting the information of
The property search support system according to claim 1, characterized in that. The storage device is
A pattern dictionary that defines normalization rules for descriptions in natural language, a contradiction resolution rule that defines unique rules for inconsistent expressions in descriptions in natural language, and a conversion dictionary that defines conversion rules for search conditions for descriptions in natural language. Hold further,
The arithmetic unit is
The process of applying the description regarding the condition of the purchase information received from the customer to the pattern dictionary to normalize and extracting the meaning included in the description, and the above-mentioned contradiction resolution rule for the pair in which the meanings of the extracted meanings are inconsistent with each other. , And the process of narrowing down to one of the meanings of the pair, and the process of applying the meaning obtained by the extraction and the meaning unique by the narrowing down to the conversion dictionary and converting it into a search condition. Do more,
The search condition obtained by the conversion is used as the condition for the purchase information, and the condition for the purchase information and the condition for the unsuccessful sale information held in the storage device are applied to the contract probability model to obtain the sale. It further executes a process of determining the contract probability of information and outputting the information of the contract probability to a predetermined device.
The property search support system according to claim 1, characterized in that. Information processing system
In the storage device, the sale information, purchase information, and contract information of the real estate property are retained,
Processing that uses the pair of selling information and buying information indicated by the contract information as regular learning data,
Processing to specify the range of difference between the desired conditions indicated by the purchase information in the contract information and the conditions of the actual contracted real estate property,
Processing in which the desired condition is included in the difference range and the pair of buy information and sell information in which the sale is unsuccessful is used as negative learning data.
A process of executing machine learning based on the positive example learning data and the negative example learning data to generate a contract probability model for real estate sales.
Property search support method to execute. The information processing system
A plurality of contract probability models are generated, and one contract probability model is generated by performing predetermined statistical processing on the plurality of contract probability models.
The property search support method according to claim 5, characterized in that. The information processing system
The contract probability of the sold information is determined by applying the condition of the buy information received from the customer and the condition of the unsuccessful sale information held in the storage device to the contract probability model, and the contract probability is determined. Further executes the process of outputting the information of
The property search support method according to claim 5, characterized in that. The information processing system
In the storage device, a pattern dictionary that defines a normalization rule for a description in natural language, a contradiction resolution rule that defines a unique rule for a contradiction expression in a description in natural language, and a conversion rule for a search condition for a description in natural language are provided. Keeps the specified conversion dictionary further,
The process of applying the description regarding the condition of the purchase information received from the customer to the pattern dictionary to normalize and extracting the meaning included in the description, and the above-mentioned contradiction resolution rule for the pair in which the meanings of the extracted meanings are inconsistent with each other. , And the process of narrowing down to one of the meanings of the pair, and the process of applying the meaning obtained by the extraction and the meaning unique by the narrowing down to the conversion dictionary and converting it into a search condition. Do more,
The search condition obtained by the conversion is used as the condition for the purchase information, and the condition for the purchase information and the condition for the unsuccessful sale information held in the storage device are applied to the contract probability model to obtain the sale. Further executing a process of determining the contract probability of information and outputting the information of the contract probability to a predetermined device.
The property search support method according to claim 5, characterized in that.

Images