A hotel check-in management system based on intelligent interaction

Abstract

The application discloses a hotel check-in management system based on intelligent interaction and relates to the technical field of hotel management.The system comprises a data acquisition module, a multi-dimensional matching module, a relationship graph acquisition module, a work-rest inference module, a compatibility analysis module and a score correction module.The data acquisition module acquires a passenger's accommodation intention, a travel purpose, a dynamic environment portrait of a room and a work-rest state of an assigned room.The multi-dimensional matching module calculates an original comprehensive matching score.The relationship graph acquisition module acquires a room adjacency relationship graph representing acoustic coupling strength.The work-rest inference module infers a passenger's work-rest type probability distribution according to the travel purpose.The compatibility analysis module extracts assigned neighbors of a candidate room, combines the work-rest distribution and the neighbor work-rest state to calculate an expected conflict degree, and obtains an adjacent conflict risk index by weighting with edge weights, and then maps the adjacent conflict risk index into an adjacent compatibility correction coefficient.The score correction module adjusts the original score with the correction coefficient.The recommendation output module sorts the recommended rooms according to the corrected scores and outputs the recommended rooms in combination with hard constraints.The application can effectively reduce noise conflicts between adjacent rooms and improve the check-in experience.

Description

Technical Field

[0001] This invention relates to the field of hotel management technology, specifically to a hotel check-in management system based on intelligent interaction. Background Technology

[0002] In the field of intelligent hotel management technology, room check-in management and intelligent allocation are key links to improve the guest accommodation experience and hotel operation efficiency. With the development of intelligent technology, room allocation systems are widely used in the scenario of guest check-in, aiming to automatically recommend the most suitable candidate rooms based on the guest's personalized needs, combined with the physical attributes of the room and real-time environmental data.

[0003] Existing smart room allocation systems typically treat environmental parameters such as the quietness of each room as inherent attributes independent of the allocation decision. Their design is based on static tags in the rooms or real-time readings from independent sensors in each room, calculating the optimal room matching score for each guest independently and allocating rooms accordingly.

[0004] However, this allocation method suffers from systemic failure when multiple guests check into a hotel simultaneously. Because of the acoustic coupling channels between adjacent guest rooms in the hotel building's physical structure, the actual quietness of a guest room is dynamically affected by the allocation status of adjacent rooms. Existing technologies employ an independent matching decision mechanism, completely ignoring the spatial transmission of noise between adjacent rooms and the conflicting sleep patterns of different guests. It is precisely this technical feature of treating guest room quietness as an independent attribute for independent matching that leads to multiple guests with incompatible sleep patterns being assigned to adjacent rooms. This causes the quietness of the system-recommended guest room to be severely degraded after the adjacent rooms are occupied, resulting in a sudden failure where the recommended quiet room becomes noisy after occupancy.

[0005] Therefore, a hotel check-in management method based on intelligent interaction is proposed. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a hotel check-in management system based on intelligent interaction.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows: In a first aspect, the present invention discloses a hotel check-in management system based on intelligent interaction, comprising: The data acquisition module is used to acquire the accommodation intention data and travel purpose tags of the passengers currently to be assigned, as well as the dynamic environmental profiles of each candidate room and the rest and activity status information of the guests in the assigned rooms; The multi-dimensional matching module is used to perform multi-dimensional independent matching based on accommodation intent data and dynamic environmental profiles, and calculate the original comprehensive matching score of each candidate room. The relationship graph acquisition module is used to acquire a pre-built guest room adjacency relationship graph. The guest room adjacency relationship graph has hotel guest rooms as vertices and physical adjacency relationships as edges, and the edge weights represent the acoustic coupling strength between adjacent guest rooms. The daily routine inference module is used to infer the probability distribution of the daily routine type of the current passenger to be assigned based on the travel destination label; The compatibility analysis module is used to extract the assigned neighbors of each candidate room in the room adjacency graph. If there are assigned neighbors, the expected conflict level is calculated based on the guest activity status information and activity type probability distribution of the assigned neighbors. The expected conflict level is weighted by combining the edge weight between the assigned neighbors and the candidate room to obtain the adjacency conflict risk index. The adjacency conflict risk index is then mapped to the adjacency compatibility correction coefficient. The score correction module is used to correct the original comprehensive matching score using the adjacency compatibility correction coefficient, so as to obtain the corrected comprehensive matching score. The recommendation output module is used to sort each candidate room according to the corrected comprehensive matching score, and output the final room recommendation result by combining the preset hard constraint check conditions.

[0008] Secondly, this invention discloses a hotel check-in management method based on intelligent interaction, comprising the following steps: Obtain accommodation intention data and travel purpose tags of passengers currently awaiting allocation, as well as dynamic environmental profiles of each candidate room and information on the daily routines of guests in already allocated rooms; Based on accommodation intent data and dynamic environmental profiles, multi-dimensional independent matching is performed to calculate the original comprehensive matching score for each candidate room. Obtain a pre-constructed guest room adjacency graph, which uses hotel guest rooms as vertices and physical adjacency relationships as edges, with edge weights representing the acoustic coupling strength between adjacent guest rooms; Infer the probability distribution of the daily routine type of the current passengers to be assigned based on the travel purpose label; For each candidate room, its assigned neighbors in the room adjacency graph are extracted. If there are assigned neighbors, the expected conflict level is calculated based on the guest activity status information and activity type probability distribution of the assigned neighbors. The expected conflict level is then weighted by the edge weight between the assigned neighbors and the candidate room to obtain the adjacency conflict risk index. The adjacency conflict risk index is then mapped to the adjacency compatibility correction coefficient. The original comprehensive matching score is corrected using the adjacency compatibility correction coefficient to obtain the corrected comprehensive matching score; The candidate guest rooms are sorted according to their comprehensive matching scores, and the final guest room recommendation results are output based on the preset hard constraint check conditions.

[0009] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By obtaining the adjacent relationship graph of guest rooms and edge weights (representing the acoustic coupling strength), and calculating the adjacent conflict risk index based on the probability distribution of the guest's daily routine status and the current guest's daily routine type, the adjacent compatibility correction coefficient is used to correct the original matching score, so that the recommendation results can avoid the risk of adjacent noise conflict in advance, significantly reducing the room change rate and complaint rate caused by adjacent noise, and ensuring the actual check-in experience of guests. 2. By assigning differentiated acoustic coupling weights (maximum for shared walls, minimum for diagonal offsets) to each edge in the guest room adjacency diagram based on physical adjacency conditions such as shared walls, shared floors, and diagonal offsets, and by performing weighted averaging when calculating the adjacency conflict risk index, the system can truly reflect the intensity differences of noise transmission in different directions, avoid the one-size-fits-all evaluation method for all neighbors, and improve the physical accuracy of adjacency compatibility correction. 3. Based on the revised comprehensive matching score ranking, room availability constraints, accessibility compliance constraints, and fire evacuation compliance constraints are further superimposed to ensure that the recommendation results conform to the actual operation rules of the hotel. At the same time, after the guest confirms acceptance, the probability distribution of the rest time type is written into the allocated record cache for real-time reference in the adjacent coupling analysis of subsequent concurrent guests. This realizes the global dynamic linkage of room status in multi-guest concurrent scenarios, improving the overall efficiency and resource utilization of occupancy allocation during peak periods. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is an overall block diagram of the system according to Embodiment 1 of the present invention; Figure 2 This is a timing diagram of a concurrent check-in scenario according to Embodiment 1 of the present invention; Figure 3 This is an overall block diagram of the method in Embodiment 2 of the present invention. Detailed Implementation

[0012] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Example 1:

[0014] like Figures 1-2 As shown, a hotel check-in management system based on intelligent interaction includes: For ease of understanding, the following will use a building with Let's take a business hotel with a few guest rooms as an example. Suppose on a certain day... to During peak check-in periods, the system receives There are 10 concurrent check-in requests from passengers awaiting assignment, of which the current passengers awaiting assignment are passengers. Passengers The traveler's intended accommodation, entered via mobile device, was described as "I'm a light sleeper and would like a room away from the elevator, not facing the street, and as quiet as possible." The order information indicated the traveler's travel purpose was tagged as "business trip," and the estimated check-in time was [date missing]. Check-out time is expected to be the following day. The hotel currently has some rooms in a "occupied" or "assigned but not yet occupied" status, for example... The rooms have been allocated to group travelers. The room is already occupied by families on vacation. For business travelers who have already checked into the room, the subsequent handling process can be explained using this specific scenario.

[0015] The data acquisition module is used to acquire the accommodation intention data and travel purpose tags of the passengers currently to be assigned, as well as the dynamic environmental profiles of each candidate room and the rest and activity status information of the guests in the assigned rooms; This module forms the foundational data for subsequent room matching and adjacency coupling analysis. The accommodation intent data of guests currently awaiting allocation can come from self-service check-in terminals, mobile applications, or front desk interfaces; it can be text directly entered by the guest or standardized text obtained through speech recognition. For speech input, the system first converts it to text using a standard speech recognition module, then performs standardization processing such as removing colloquialisms, unifying synonyms, and correcting obvious recognition errors. For example, a guest... The original voice message, "I'm a light sleeper, don't assign me a room near the elevator or facing the street, preferably somewhere quiet," can be standardized to generate accommodation intent data. "I'm a light sleeper, so I'd like a room that's away from the elevator, not on the street, and as quiet as possible."

[0016] Trip destination tags This information can be obtained from order information and passenger profiles, and can include categories such as business trips, family travel, leisure vacations, group travel, conference attendance, and others. The system can determine the travel purpose tag based on information such as travel reimbursement type, number of accompanying persons and age structure, booking channel, and conference / group room identification. Using the example above, the passenger... The order includes a corporate rate and travel expense reimbursement markings, and is for a single occupancy, therefore the system will classify the traveler as a single person. The travel purpose tag is determined to be business trip. If the order contains... Adults and If there are 10 children and the booking channel is a parent-child travel channel, the travel purpose label can be identified as family travel; if the order belongs to a batch booking under the same group code, the travel purpose label can be identified as group travel.

[0017] During the same peak check-in period, the system can also obtain a set of concurrent passengers awaiting assignment. The data for each passenger may include standardized passenger request text. Trip destination tags Expected check-in time Expected check-out time and passenger history complaint records To ensure that the room availability and schedules of previously assigned guests can be used by subsequent guests, the system can be configured according to the estimated check-in time. right Sort in ascending order; when multiple travelers have the same expected check-in time, priority is given to those whose orders were created earlier. Continuing with the above scenario, if travelers... Expected Check-in, guests Expected Check-in, guests Expected Check-in and guests The order was created earlier than the passenger's. The processing order is: passenger Passengers Passengers In this way, passengers Once the room number and guest's daily routine information are assigned, they can be used as a reference for travelers. Allocated guest room information during adjacent coupling analysis.

[0018] The dynamic environmental profile of each candidate guest room can be constructed by combining the room's static attributes with real-time sensor data. The room's static attributes are derived from the master data of the hotel property management system and include at least each guest room. Floor number Orientation label ,area and room type tags Real-time sensor data comes from a distributed IoT sensor network within or around hotel rooms, and includes at least real-time sound pressure levels. Real-time illuminance and real-time temperature Among them, real-time sound pressure level The current method can be adopted. A-weighted equivalent continuous sound pressure level within a minute time window, real-time illuminance Reflects the current lighting or blackout status of the guest room, and the real-time temperature. It can be retained as comfort monitoring data.

[0019] In its implementation, the system reads each guest room from the hotel property management system. Current room status marker Room status marker Possible values Only when At that time, guest room Enter the pool of candidate rooms. For each candidate room... System construction and matching dimension set Aligned dynamic environment profile .in, This is the current value for the quietness dimension, taken from the real-time sound pressure level. ; This is the current value for the lighting dimension, taken from real-time illuminance. ; This is the current value for the floor dimension, taken from the floor number. ; The current value for the spatial dimension is taken from the area. ; The current value for the convenience dimension is taken from the guest room. Straight-line distance from the nearest elevator shaft ; The current value for the landscape dimension is taken from the hotel's pre-defined landscape score. .

[0020] For example, in the aforementioned hotels, candidate rooms The real-time sound pressure level of the room is Real-time illuminance is Floor number is The area is The distance to the nearest elevator shaft is The landscape rating is This can create a dynamic environmental profile. Candidate Rooms The real-time sound pressure level of the room is Real-time illuminance is Floor number is The area is The distance to the nearest elevator shaft is The landscape rating is This can create a dynamic environmental profile. These dynamic environmental profiles can directly serve subsequent multi-dimensional independent matching, enabling the system to not only rely on static room type labels, but also utilize real-time acoustic and illuminance conditions.

[0021] If real-time sensor data for a candidate room is missing or faulty, the system does not discard the candidate room directly. Instead, it uses the average value of the group of rooms on the same floor and facing the same direction as the candidate room during the same historical period as a substitute value, and marks the data confidence level as downgraded in the dynamic environment profile of the candidate room. For example, if Real-time sound pressure level of the room It is a null value, and Rooms on the same floor facing the same direction are near sky to The average sound pressure level during the time period is Then As Replace the values ​​and set the data confidence level flags. Guest room settings for normal acquisition of real-time sensor data This process avoids making candidate rooms completely unusable due to short-term anomalies in a single sensor, while also providing a basis for penalty processing in the subsequent raw composite matching score through confidence labeling.

[0022] The guest activity status information for assigned rooms can be retrieved from the room status records and assigned record cache of the hotel property management system. For guests whose rooms have been assigned, the system saves their assigned room number, travel purpose tag, activity type probability distribution, and confirmed activity type tag. The confirmed activity type tag can be the type with the highest probability in the guest's activity type probability distribution. For example, The room has been allocated to a group traveler. If this traveler is subsequently inferred to be most active at night, then... The room's guests' daily routine information can be recorded as nighttime active; The room is already occupied by families traveling together, with families with children being the most likely group. The room's occupant's daily routine information can be recorded as "family / parent-child". For currently unassigned passengers... In this regard, the information on the daily routines of guests in these assigned rooms will not be used to change the static attributes of the rooms, but rather to determine whether there is a risk of adjacent noise coupling between the candidate rooms and the surrounding assigned rooms that could cause sudden changes in quietness.

[0023] Through the data acquisition process described above, the system simultaneously obtains three types of complementary information: First, traveler-side accommodation intention data and travel purpose tags, used to characterize what the currently awaiting traveler "wants" and "may have certain daily routine characteristics"; second, dynamic environmental profiles of the guest rooms, used to characterize the current perceptible environmental state of each candidate guest room; and third, information on the daily routines of guests in already assigned guest rooms, used to characterize existing or soon-to-be-existing behavioral noise sources in adjacent rooms. Subsequently, based on this information, the system can calculate the original comprehensive matching score between each candidate guest room and the traveler's intention, and further refine the recommendation priority of the candidate guest rooms by incorporating the daily routines of already assigned neighbors.

[0024] The multi-dimensional matching module is used to perform multi-dimensional independent matching based on accommodation intent data and dynamic environmental profiles, and calculate the original comprehensive matching score of each candidate room. In this module, the system performs semantic parsing and quantification of the accommodation intent data of the passengers currently awaiting allocation. Then, it independently compares this data with the dynamic environmental profiles of each candidate room across multiple matching dimensions. Finally, the comparison results from each dimension are aggregated into a comprehensive value based on passenger preference weights. The input to this module is the passenger's... Accommodation intention data (i.e., "I'm a light sleeper and want a room that's away from the elevator, not facing the street, and as quiet as possible"), travel purpose tag (Business trips), historical complaint records (Including one historical complaint, the complaint type is "noise"), and a set of dynamic environmental profile vectors for each candidate room.

[0025] First, regarding accommodation intention data The system performs semantic entity recognition, extracts preference intensity coefficients for each matching dimension, and normalizes them to generate a traveler intent preference weight vector. The system uses a pre-built semantic entity dictionary in the hotel domain as the matching benchmark. Keyword matching and dependency parsing are performed to categorize semantic entities into six matching dimensions—dimensions Quietness, Dimensions Lighting, Dimensions Floors, Dimensions Space, dimension Convenience, dimensions Landscape. Continuing with the example above, travelers... The input text is parsed into dimensions. The three sub-intents under the quietness level are "away from the elevator," "not facing the street," and "quiet," but none of the five dimensions are described. It should be noted that although "away from the elevator" involves the convenience dimension, the traveler's intention in this context is to obtain a quiet environment by moving away from the elevator. Therefore, the system classifies this expression as the quietness dimension rather than the convenience dimension. This context-sensitive classification can be determined through the modification relationship of "want...away from the elevator...quiet" in dependency syntax.

[0026] Subsequently, the system quantifies the semantic intent strength of each dimension into a dimension preference strength coefficient. Passengers In dimensions There are several clear positive statements regarding quietness. There is no description of the other five dimensions. .

[0027] For dimensions where no preference was expressed, the system uses travel purpose tags. Prior weight matrix from intention to physical parameters The prior weights of the group preferences corresponding to the travel destination are loaded to complete the calculation. for 3D matrix, elements Indicates the category of travel purpose The passenger group in the dimension The average preference weights on the data. For business travelers, the group prior preference weights can be set as follows: (Quietness is a key factor for business travelers.) (lighting), (floor), (space), (Convenience) (Landscape). Regarding The dimension that makes ,Right now , , , , .

[0028] Next, the system checks the historical complaint record set. Passengers The file contains a historical record of a complaint type "noise," corresponding to the dimension. Quietness. The system will consider dimensions. The preference intensity coefficient increased to If the passenger does not explicitly express a preference for quiet in the text (i.e., Originally for If so, the historical complaint mechanism will... Upgraded to This ensures that the quiet preferences of travelers who have previously complained about noise are not overlooked.

[0029] Finally, for all The preference intensity coefficients of each dimension are normalized to generate a passenger intention preference weight vector. The normalization formula is:

[0030] ; Continue to use passengers For example, the sum of the preference intensity coefficients for each dimension is After normalization, we get , It can be seen that the weight of the quietness dimension exceeds [a certain percentage]. This reflects the passenger's strong need for quiet, which stems both from the explicit statement in the text and from the reinforcement of historical complaint records.

[0031] Based on this, the system simultaneously maps accommodation intent data into target parameter vectors for each matching dimension. The mapping rules are based on a semantic-physical parameter lookup table for the hotel domain.

[0032] Dimension Quietness: Passengers The system maps the semantic entity containing emphatic quiet expressions such as "light sleep" to the target sound pressure level. If it only includes general expressions of quiet such as "quiet" or "not noisy," then it is mapped to... If there is no description related to quietness, then take... As a lenient threshold.

[0033] Dimension Lighting: Passengers No lighting preferences were expressed; the target value for this dimension was the median illuminance of all guest rooms. Using the aforementioned hotel, the median real-time illuminance of 200 guest rooms was approximately... ,make Furthermore, due to passengers Expected check-in time It falls within the nighttime check-in period. Even if the passenger subsequently expresses the meaning of "good lighting," the system will automatically switch the meaning to "good shade," mapping it to... Extremely low light target. But current passengers Since no lighting preference was expressed, the day / night semantic discrimination logic was not triggered.

[0034] Dimension Floor to Dimension Landscape: Tourists No preference was expressed, and the target value was taken as the median of its respective dimension. (Mid-floor number) (Median area) (Distance from the middle elevator) (Median landscape score).

[0035] Next, a dynamic environmental profile is constructed based on the static attributes of each candidate guest room and real-time sensor data. The status of each room is then tagged. For guest rooms, the system integrates their static attributes with real-time sensor data to construct a dynamic environmental profile vector aligned with six matching dimensions. . Real-time readings from acoustic sensors , Real-time readings from the light sensor , Taken from PMS static properties , Taken from PMS static properties , Pick up guest rooms Straight-line distance from the nearest elevator shaft , Take the hotel's pre-defined view rating .

[0036] For guest rooms with missing or faulty sensor data, the system uses the group of rooms on the same floor and facing the same direction as the candidate guest room to... The historical average value for the same time period is used as a proxy value, and the data confidence level is downgraded in the dynamic environmental profile of the candidate room. For example, if The room's acoustic sensor readings were blank, while nearby guest rooms on the same floor facing the same direction... sky to The average sound pressure level during the time period is Then As an alternative and set Guest room tags for normal data acquisition .

[0037] Using the aforementioned hotel example, candidate rooms The dynamic environmental profile of the house is Candidate rooms The dynamic environmental profile of the house is Candidate rooms The dynamic environmental profile of the room (using a substitute value for sound pressure level) is as follows: ,and .

[0038] For each matching dimension, calculate the single-dimensional matching score between the current state value in the dynamic environment profile of each candidate guest room and the target parameter of the corresponding dimension. The range of values ​​is , This indicates a perfect match. The system employs differentiated matching functions based on different dimensions of physical characteristics.

[0039] For dimensions Quietness, belonging to the "lower is better" dimension of noise: when the real-time sound pressure level of the guest room is lower than or equal to the target sound pressure level, the single-dimensional matching score is full; when the real-time sound pressure level of the guest room exceeds the target sound pressure level, the single-dimensional matching score decreases linearly with the increase of the excess, until it reaches the complete mismatch range and drops to zero. Single-dimensional matching score for the quietness dimension. The specific calculation formula is as follows:

[0040] ; in The real-time sound pressure level of the candidate guest rooms. The target sound pressure level for the quietness dimension. This represents the complete mismatch range for the quietness dimension, with the default value. This indicates that the sound pressure level in the guest room exceeds the target value. The matching score drops to zero. The physical meaning of this formula is: if the current sound pressure level of the guest room is lower than the guest's target threshold, it means that the guest room fully meets the requirements in terms of quietness, and the score is full marks no matter how good the actual quietness is; if it exceeds the target threshold, the more it exceeds, the lower the score, reflecting the increasing marginal impact of noise exceeding the standard.

[0041] Continue to use passengers Target parameters , Real-time sound pressure level of the room Exceeding the target value Substituting into the formula, we get The room's quietness level is close to travelers' expectations, earning a high score. Real-time sound pressure level of the room Exceeding the target value Substituting into the formula, we get point. Room sound pressure level (Alternative value), exceeding the target value Substituting into the formula, we get point.

[0042] For dimensions Lighting, which falls under the "closer to the target, the better" dimension of target matching, has a matching score of:

[0043] ; in The real-time illuminance of the candidate guest rooms. For the target illuminance, This represents the complete mismatch range in terms of lighting dimensions, with the default value. Passengers No preference for natural light was expressed. . Room illumination is ,deviation , point; Room illumination is ,deviation , point.

[0044] For dimensions Floor level, which also belongs to the target matching dimension, has a matching score as follows:

[0045] ; in The floor number for the candidate guest rooms. For the target value of the floor, This represents the complete mismatch range for the floor level dimension, with the default value. Passengers No floor preference expressed. All three candidate guest rooms are located in Floor, floor deviation , The score is [score missing]. It should be noted that this lower score does not come from the travelers' active requests, but from a neutral benchmark value for completing the travel purpose. It has the same impact on the three candidate rooms and will not change their ranking.

[0046] For dimensions Space, which belongs to the "the bigger the better" dimension, has a matching score of:

[0047] ; in The area of ​​the candidate guest rooms. This is the minimum target area. Full marks are awarded when the guest room area meets or exceeds the target; otherwise, the score is reduced proportionally. (Guest) Spatial preference not expressed . Room area , Score (exceeding the target value, full marks); Room area , point.

[0048] For dimensions Convenience, which falls under the "shorter distance, better" dimension, has a matching score of:

[0049] ; in This is the straight-line distance from the guest room to the nearest elevator shaft. The target value is the maximum distance to the elevator. The default value represents the complete mismatch range for the convenience dimension. Passengers No convenience preference was expressed. (Neutral baseline, i.e., the median elevator distance for all guest rooms). Since the elevator distances for all three candidate guest rooms do not exceed [a certain value]... ( house , house , house All three rooms did not exceed the target value, therefore all three rooms received full marks in this dimension. point.

[0050] For dimensions Landscape, which falls under the "higher the score, the better" dimension, has the following matching score:

[0051] ; in Rate the view of the hotel according to the pre-defined criteria. This represents the target value for the landscape rating. (For tourists) No landscape preference expressed. . House view rating , point; House view rating , point.

[0052] Summarize the single-dimensional matching scores for each of the above dimensions. The dimension score vector of the house is , The dimension score vector of the house is .

[0053] After calculating the single-dimensional matching scores for each dimension, the system uses the passenger intent preference weight vector to perform a weighted summation of the scores for each dimension to obtain the original comprehensive matching score. The weighted summation formula is:

[0054] ; in For travelers in dimensions Intentional preference weights, For guest rooms In dimensions The single-dimensional matching score is used. (This is based on the passenger...) Preference weights , The house's original overall matching score is: ; The house's original overall matching score is: .

[0055] As you can see, The house's original score was higher than The main reason for choosing a room is that the quietness dimension has the highest weight (exceeding the previous one). ),and The house's score on this dimension ( (points) significantly better than house( (Points). This indicates that in purely independent matching mode, the system will prioritize recommendations. house.

[0056] For guest rooms with downgraded data confidence, the system multiplies the weighted sum by a preset confidence penalty factor to reduce their original comprehensive matching score. Continuing with the example above, The sound pressure level of the room uses a substitute value ( ), and its single-dimensional matching scores for each dimension are The weighted sum score is Multiply by the confidence penalty factor Finally, the original comprehensive matching score The penalty mechanism is based on the following: when sensor data is missing, the average deviation between the historical average substitute value and the actual value under normal operating conditions is approximately [missing value]. - , The penalty coefficient can cover this deviation range.

[0057] Through the aforementioned multi-dimensional independent matching and weighted aggregation process, the system generates an original comprehensive matching score for each candidate room. This score represents the degree of matching between each candidate room and the current traveler's intent, without considering the allocation status of adjacent rooms. Continuing with the example above, under the condition of considering only independent matching... Room division point, Room division point, Room division However, this ranking does not take into account the noise coupling effect between adjacent guest rooms. For example, Although Fang scored the highest in independent matches, its left-hand neighbor... The rooms have been allocated to nighttime-active group travelers, and this adjacent arrangement may lead to... The room experienced a sudden change in quietness after occupancy. These original scores will be corrected through adjacency coupling analysis.

[0058] The relationship graph acquisition module is used to acquire a pre-built guest room adjacency relationship graph. The guest room adjacency relationship graph has hotel guest rooms as vertices and physical adjacency relationships as edges, and the edge weights represent the acoustic coupling strength between adjacent guest rooms. The guest room adjacency diagram is built once during system deployment and can be reused in subsequent allocation decisions without repeated calculations. Its construction is based on hotel building topology data, typically derived from CAD or BIM models created during the hotel's construction phase. This data accurately records the physical location coordinates, floor information, and structural relationships of each guest room with adjacent spaces.

[0059] In the specific construction process, the system uses all hotel rooms as the vertex set. The set of edges consists of pairs of guest rooms that satisfy the condition of physical adjacency. Construct an undirected weighted graph The term "physical adjacency" refers to the existence of a direct acoustic coupling path between two guest rooms through the building structure, meaning noise can be transmitted from one guest room to another via shared building components. The system checks each pair of guest rooms one by one from the building topology data to see if it meets one of the following three adjacency conditions: Condition A (Shared Wall): Two guest rooms are located on the same floor and have consecutive room numbers, sharing a partition wall. For example... House and The house is located at the The floors are adjacent in number and share a partition wall. This wall may contain pipe shafts or electrical conduits, creating a direct acoustic transmission channel. For example... House and The rooms also meet the condition of sharing a wall. Sharing a wall is the most significant way of noise transmission in hotel buildings because the partition wall has a large area and limited mass, and the pipes and cable trays inside the wall may form sound bridges.

[0060] Condition B (Shared Floor): Two guest rooms are located on adjacent floors and are horizontally aligned vertically, sharing a floor or ceiling. For example... House and The rooms are located in the first Layer and first layer, The floor of the room is exactly The ceiling of the room can transmit sounds such as footsteps, furniture dragging, and other impact noises, as well as airborne noises like television broadcasts and conversations, all through the floor slab. This is especially noticeable between rooms with adjacent floors where children are staying, as the impact noise from toddlers playing on the floor significantly affects the rooms below.

[0061] Condition C (Diagonal Offset): Two guest rooms are located on adjacent floors but are horizontally offset by one room, with an oblique acoustic transmission channel between them. For example... House and Although the characters for "room" and "room" do not directly correspond to each other in terms of height, House and Shared walls Fang Youyu In buildings with shared floors, noise can be indirectly transmitted along either a "wall → floor" or a combination of "floor → wall" paths. The diagonally offset path, due to the additional attenuation from passing through a building component, has a significantly lower coupling strength than the first two directly adjacent types.

[0062] When two guest rooms and When any of the above conditions are met, the system is in the edge set China establishes border It should be noted that the same pair of guest rooms cannot simultaneously meet two different adjacent conditions (e.g., House and (A room only needs to meet the condition of sharing a wall, not the condition of sharing a floor slab), therefore each edge corresponds to only one type of adjacency. This construction process is performed throughout the entire hotel, with one building as the foundation. Guest rooms, Taking a multi-story hotel as an example, the average revenue per room is approximately to Strip edge (shared wall on both sides) Strip, above the floor slab Strip, below the common floor slab Strip, diagonal offset to (strips), the entire image contains approximately to Edge.

[0063] Using the hotel example mentioned above, The neighbors of the room in the guest room adjacency diagram include: Rooms (sharing a wall, adjacent on the left), Rooms (sharing a wall, adjacent on the right), Room (with a total floor slab, directly below) Room (with a total floor slab, directly above) Room (diagonally offset, lower left) Room (diagonally offset, lower right) Room (diagonally offset, upper left) Room (diagonally offset, upper right), total Each neighboring node.

[0064] While establishing edges, the system assigns acoustic coupling weights to each edge based on its adjacency type. This is used to characterize the relative intensity of noise as it propagates through this edge. Acoustic coupling weight. The calculation formula is:

[0065] ; in This is the baseline value for the shared-wall coupling weight. The benchmark value for the coupling weight of the shared floor slab. The diagonal offset coupling weight baseline value, and In this embodiment, , , These values ​​are based on engineering calibrations of sound insulation in building acoustics: the sound insulation of a shared wall is typically... to The total floor slab is to (Floor slabs are typically heavier than partition walls, therefore their sound insulation is slightly higher). The diagonally offset path experiences an additional wall attenuation, resulting in a total sound insulation of approximately [missing information]. to Normalization is performed based on the strongest coupling along the shared wall. Afterwards, the coupling strength of the shared floor slab is approximately equal to that of the shared wall. (Right now The coupling strength of the diagonal offset is approximately equal to that of the shared wall. (Right now ).

[0066] It should be noted that the above baseline values ​​are the default parameters during system deployment, and hotels can adjust them according to the sound insulation level of their own buildings. For example, for hotels built to high sound insulation standards (STC level of partition walls)... ), can be appropriately reduced to And adjust other benchmark values ​​proportionally; for budget hotels (STC rating) to The default value can be retained or adjusted appropriately to reflect lower sound insulation performance.

[0067] Following the example above, House and The edge between rooms Satisfying the shared wall condition, acoustic coupling weight ; House and The edges between rooms meet the requirement of sharing a floor slab. ; House and The edges between the rooms satisfy the diagonal offset condition. These weight values ​​will be used directly in the subsequent calculation of the neighbor conflict risk index, to perform a weighted average of the conflict levels of different neighbors based on acoustic coupling strength.

[0068] In addition, after the guest room adjacency diagram is constructed, the system also needs to define the schedule conflict matrix. The matrix is A 3D symmetric matrix used to quantify the degree of conflict between different sleep patterns. Matrix elements. The range of values ​​is ,in Indicates full compatibility. This indicates a serious conflict. Based on statistical patterns of guest behavior and routines in the hotel operations sector, The specific values ​​are:

[0069] ; Each row (column) corresponds to the following: =Early to bed and early to rise type =Late-night, late-waking-up type =Night-active type =Family / Parent-Friendly Type. The values ​​in this matrix reflect the actual noise interference relationships between travelers with different sleep patterns in a hotel accommodation setting. For example, the conflict value between early-to-bed, early-to-rise travelers and nighttime active travelers. This is because nighttime active travelers are... Subsequent noise from activities (such as chatting, watching TV, and using the bathroom) directly intrudes on the rest time of early-to-sleep travelers, causing the most severe conflict. The conflict value between late-to-sleep, late-to-rise travelers and family-oriented travelers is also significant. This is because young children usually wake up in the early morning. They had already woken up and were making noise, which happens to be the time when night owls are in deep sleep. The conflict between early-to-bed, early-to-rise travelers and family-oriented travelers is significant. While early morning activities in toddlers do not encroach on their pre-sleep rest period, they can still affect the quality of deep sleep in the morning for early-to-bed travelers. The conflict value between early-to-bed, early-to-rise travelers and late-to-bed, late-to-rise travelers... Both only to There is slight interference during the transition period. All diagonal elements are... This indicates that there is no conflict in the schedules of passengers with the same work-rest pattern.

[0070] Guest room adjacency diagram Acoustic coupling weights on each side and the work-rest conflict matrix Together, these form the foundational data structure for adjacent room conflict risk analysis in subsequent processes. During system operation, whenever a passenger completes an assignment, their assigned room number and confirmed activity / rest type label are recorded in the assigned record cache. When processing subsequent passengers, the system can extract the assigned neighbors of candidate rooms from the room adjacency graph and, combined with the activity / rest type and acoustic coupling weight of the assigned neighbors, assess the adjacent noise coupling risk faced by the candidate room. Continuing with the above example, when a passenger... When entering processing, Rooms have been allocated to group travelers (whose activity type has been confirmed as nighttime active). ), The room is occupied by families traveling together (whose travel type has been confirmed as family-friendly). The daily routines of these assigned neighbors will directly affect Risk assessment of neighboring conflicts.

[0071] The daily routine inference module is used to infer the probability distribution of the daily routine type of the current passenger to be assigned based on the travel destination label; While the initial comprehensive matching score is calculated, the system simultaneously performs preparatory work for adjacent coupling risk analysis. The primary step is to infer the probability distribution of the traveler's daily routine type based on the travel purpose tag of the passenger to be assigned. The reason for using travel purpose instead of directly asking about the passenger's daily routine is that in actual hotel operations, passengers typically do not proactively disclose their preferences. The travel purpose tag can be obtained from multiple signals, including the travel expense reimbursement type marking in the order information, the number and age structure of accompanying persons, and the booking channel—a readily available data source. Although there is no strictly deterministic correspondence between travel purpose and daily routine type, under the statistical patterns of hotel customer demographics, passenger groups with different travel purposes exhibit identifiable clustering characteristics of daily routine patterns.

[0072] In its implementation, the system obtains a preset prior transition probability matrix from travel destination to work / rest type. The rows of this prior transition probability matrix correspond to different travel destination categories, and the columns correspond to different work / rest schedule types. The system uses the travel destination label of the passenger to be assigned as an index to retrieve the information from the prior transition probability matrix. Extract the corresponding row and use it as the probability distribution vector of the current passenger's work and rest type. Each component in the probability distribution vector of the work and rest type represents the probability that the passenger belongs to the corresponding work and rest type.

[0073] Prior transition probability matrix Based on the statistical pattern identification of groups using historical hotel operating data, including Trip (corresponding to six travel purposes) and Columns (corresponding to four sleep patterns: early to bed and early to rise, late to bed and late to rise, nighttime active, and family-oriented). Specific values ​​are as follows: When the purpose of travel is business trip, the probability distribution vector of the sleep pattern type. This indicates that the passenger has The probability is that the person is an early riser. The probability is that the person is a late sleeper and late riser. The probability is that the person is active at night. The probability is for family-oriented travel. When the purpose of the trip is family tourism. Families with children are the most likely to travel. When the purpose of the trip is leisure and vacation... Those who stay up late and wake up late are most likely to be this type of person. When the purpose of the trip is a group trip... Nighttime activity is the most likely. When the purpose of travel is to attend a conference... Those who go to bed early and wake up early are the most likely to be early risers. When the purpose of travel is other than... The probabilities of the four sleep patterns are relatively even.

[0074] It should be noted that this prior transition probability matrix The calibration process is as follows: Hotels can extract historical occupancy data from the PMS system for the past year, and combine this data with guest travel purpose tags and actual behavior data during their stay (such as self-check-out time, breakfast time, and late-night room activity sensor trigger records) to calculate the distribution ratio of different activity patterns under each travel purpose category. If the hotel lacks sufficient historical data, it can directly use the above default matrix values, which are based on a medium-sized business hotel (200 rooms, average annual occupancy rate...). to The statistical regularity of the customer source structure is determined, which has approximately [missing information] under normal operating conditions. to The classification accuracy is sufficient to produce an effective allocation correction effect.

[0075] Using the aforementioned hotel example, travelers Travel destination tags For business trips, the system transitions from the prior probability matrix. Extract the rows corresponding to business trips to obtain the passenger information. probability distribution vector of daily routine type This means passengers Most likely belongs to the early-to-bed, early-to-rise type. (probability), which determines the work-rest type label The type with the highest probability in this probability distribution vector is the early-to-bed, early-to-rise type. ).

[0076] This probability distribution-based, rather than binary, design has significant robustness implications. When there is uncertainty regarding travelers' lifestyle patterns (e.g., the probability distribution of lifestyle patterns for leisure travelers is relatively dispersed),... (Since all four types have non-zero probabilities), the system does not simply categorize them into a single lifestyle type before making a conflict determination. Instead, it retains complete probability distribution information and uses probability as the weight to perform a weighted average of all possible conflict values ​​when calculating the expected conflict level. This approach ensures that the impact of uncertain travelers on conflict risk assessment is relatively mild, avoiding excessive downgrading or upgrading due to misjudgment, and also preventing the complete disregard of a traveler's lifestyle characteristics due to uncertainty.

[0077] In the prior transition probability matrix The probability distribution for travel purposes marked "Other" is as follows: The four types have relatively even probabilities, which means that for travelers who cannot clearly determine their travel purpose, the system adopts a conservative strategy—not classifying them with a high probability into any single activity type, thereby generating a near-neutral expected conflict value in subsequent conflict assessments and avoiding allocation bias caused by insufficient information.

[0078] Furthermore, once hotels have accumulated sufficient historical data, they can adjust the prior transition probability matrix based on different seasons and customer demographics (e.g., a higher proportion of leisure and vacation guests during peak tourist seasons, and a higher proportion of conference attendees during exhibitions). Dynamic adjustments will be made to make the predicted work and rest patterns more consistent with the current actual customer composition.

[0079] The probability distribution vector of this work-rest type It will play two roles in the subsequent process: first, in processing current passengers When conducting neighbor conflict risk assessment, this vector is used in conjunction with the defined daily routine labels of assigned neighbors to calculate the expected degree of conflict between the current traveler and each assigned neighbor; secondly, when the traveler... After the allocation is completed, the work and rest type label is determined. The records will be saved along with the allocated records for future reference by travelers in their own adjacent conflict risk assessments.

[0080] Using the previous example, in handling passengers At that time, passengers who were previously at the front of the orderly queue (The purpose of the trip is group travel) The assignment has been completed, and the probability distribution of their daily routine types is as follows: The most probable type is the nocturnal active type. probability Therefore, passengers Determine your work and rest type tags (Nighttime Active) This tag has been recorded in the assigned record cache. When the system is for a passenger... Evaluate When there is a risk of conflict between neighboring houses, if Rooms have been allocated to guests The system will then use (Night-active type) This refers to the resident's daily routine information, and is related to the traveler's... Probability distribution of one's own daily routine Combine this with calculations to determine the degree of expected conflict.

[0081] The compatibility analysis module is used to extract the assigned neighbors of each candidate room in the room adjacency graph. If there are assigned neighbors, the expected conflict level is calculated based on the guest activity status information and activity type probability distribution of the assigned neighbors. The expected conflict level is then weighted by the edge weights between the assigned neighbors and the candidate room to obtain the adjacency conflict risk index. The adjacency conflict risk index is then mapped to the adjacency compatibility correction coefficient. If there are no assigned neighbors, the adjacency compatibility correction coefficient is set to the baseline value without correction. This module is the key component of the entire system that best reflects the dynamic occupancy coupling effect of adjacent rooms. Its technical essence lies in this: when multiple guests arrive successively during peak occupancy periods, the room chosen by each later arrival must not only meet their own accommodation preferences but also consider whether there are any conflicts in their sleep patterns with those of other assigned rooms. This conflict assessment is not a simple "yes or no" judgment, but rather a multi-dimensional quantitative calculation that comprehensively considers the probability distribution of the current guest's sleep patterns, the definite sleep patterns of the assigned neighbors, and the strength of the architectural acoustic coupling between the two.

[0082] In its implementation, the system targets each candidate guest room. First, extract the set of assigned neighbors for the guest room from the guest room adjacency graph. The set of assigned neighbors contains all the candidate room's neighbors in the graph that have currently been assigned to other guests, i.e., neighboring rooms marked as "assigned but not occupied" or "occupied". The system also defines an assigned neighbor existence flag. :

[0083] ; This marker is used to distinguish two distinct situations: the candidate room is surrounded by already assigned guests (…). A conflict assessment is required; or there are no assigned guests in the vicinity of the candidate room. No conflict assessment is required.

[0084] Using the aforementioned hotel example, travelers Check-in period is ,at this time Rooms have been allocated to group travelers. (Determine the work and rest type label) (i.e., nocturnal active type) The room is already occupied by family tourists. (Determine the work and rest type label) (i.e., family-oriented, parent-child type) The room is already occupied by a business traveler. (Determine the work and rest type label) (i.e., the early-to-bed, early-to-rise type). For For a room, its neighbors in the guest room adjacency diagram include: house, house, house, Houses, etc., in which neighbor groups have been allocated ,common One assigned neighbor, therefore .for For a house, among its neighbors in the adjacency diagram, if Fanghe If all rooms are already allocated, then... The allocated neighborhood group of the house is not empty. If all neighbors of a particular room (such as a room on the other side of the hotel away from the assigned area) are unassigned, then .

[0085] when When assigned neighbors exist, the system first obtains a preset conflict matrix for daily routines, where each element represents the degree of conflict between different routine types. Then, it extracts the specific routine type label for each assigned neighbor from the neighbor's resident routine status information. Next, using the routine type probability distribution vector as weight, it performs a weighted summation of the elements in the conflict matrix corresponding to the specific routine type label column to obtain the desired conflict level. Specifically, the system performs this process for each assigned neighbor... Calculate the "pair conflict expectation" between the current passenger to be assigned and the neighboring resident. The formula for calculating this value is: ; in For passengers currently awaiting assignment and their assigned neighbors The degree of conflict of expectations between them Total number of work and rest types (in this embodiment) ), The probability distribution of work and rest types The probability of each sleep pattern The first in the work-rest conflict matrix The row and assigned neighbors determine the daily routine type label. The conflict level element at the column intersection. The physical meaning of this formula is: using the probability distribution of the current traveler's activity type as weights, a weighted average is taken between the conflict values ​​of all possible combinations of activity types and the determined activity types of their neighbors, thereby obtaining a quantifiable expected conflict level while preserving the uncertainty of probability.

[0086] Continue to use passengers For example, the probability distribution of its daily routine type is as follows: .for Allocated neighbors of the house Room (guests determine their lifestyle) (i.e., nocturnal active type) travelers The process of calculating the expected value of pairwise conflict with the neighbor is as follows: From the daily routine conflict matrix Extract the first The elements of the column, i.e. Then with the passengers We take a weighted sum of the probability distributions of the daily routine types:

[0087] ; The result indicates that: due to passengers have The probability of this type of person being an early riser is high, and the conflict value between the early riser / early riser type and the nocturnal active type is as high as [missing value]. This results in a significantly higher level of expectation conflict. If passengers... If the individual is identified as an early-to-bed, early-to-rise type, then the level of conflict will be directly equal to... However, due to passengers There is still a certain probability that it belongs to other sleep patterns (such as...) The probability of being a late-to-sleep, late-to-rise type has a conflict value of only [value missing] with the nocturnal active type. Probability weighting reduces the degree of expectation conflict to .

[0088] for Another assigned neighbor of the house Room (guests determine their lifestyle) (i.e., family-oriented, parent-child type) Extract the first Column elements Calculate the expected degree of conflict:

[0089] ; for The third assigned neighbor of the house Room (guests determine their lifestyle) (i.e., the early-to-bed, early-to-rise type), from Extract the first Column elements Calculate the expected degree of conflict:

[0090] ; This value is low because The roommate is an early riser, and the guests... There is almost no conflict between the most likely sleep patterns (which are also early to bed and early to rise), only because of the traveler There is a small probability that belonging to other sleep patterns will result in a minor conflict of expectations.

[0091] After calculating the expected conflict value, the system extracts candidate guest rooms. The acoustic coupling weights of all assigned neighbors are used to calculate the neighbor conflict risk index by weighting the expected conflict level of each assigned neighbor according to its corresponding acoustic coupling weight. The specific calculation formula is as follows:

[0092] ; in Candidate guest rooms The set of assigned neighbors, Candidate guest rooms With assigned neighbors Acoustic coupling weights between them This corresponds to the expected level of conflict. The design principle of this weighted average is that different types of adjacency relationships have significant differences in noise transmission intensity—noise transmission is strongest when sharing a wall. ), adjacent to each other on the same floor slab ( ), diagonal offset adjacent weakest ( Therefore, when assessing the overall conflict risk faced by candidate guest rooms, neighbors with higher coupling strength should be given greater weight, so that neighbors with serious conflicts with the current guests' schedules and strong acoustic coupling contribute more to the risk index.

[0093] Following the example above, The acoustic coupling weights between the room and its three assigned neighbors are as follows: (Shared wall) (Shared wall) (Total floor slabs). Substitute into the formula to calculate. Neighborhood conflict risk index:

[0094] ; This result indicates that The property faces a moderate to high risk of neighborly conflicts. Fanghe Fang Yinyu The rooms share a wall (with the highest acoustic coupling weight) and are respectively classified as nighttime active type and family-oriented type (with travelers). Those who go to bed early and wake up early (both have higher conflict rates) contribute the most to the risk index; while Although the room is with the traveler While the work and rest schedules are compatible, the low coupling strength of the shared floor slab limits the overall risk mitigation effect of this compatibility.

[0095] for The house, which has been allocated to neighbors, is Room (family-friendly type) Shared wall )and Room (assuming a leisure traveler has already checked in) Determine your work and rest type That is, the type of person who goes to bed late and gets up late, with one floor slab. ). Other neighbors of the house (such as neighbors sharing a wall) Room, diagonally offset neighbor house, house, house, The houses (etc.) are currently in an unallocated state, therefore ,common There are already assigned neighbors. The system needs to first calculate the expected value of pairwise conflicts. and Same, for . The calculation is as follows, from Extract the first Column elements :

[0096] ; The risk index of neighborly conflict of the house is: ; As can be seen, although House and The room shares a wall and the degree of conflict is The rooms are the same, but due to Another neighbor of the house Room (late-night, late-wake type) and guests The conflict is low, and the coupling is only through the shared floor slab, therefore The overall risk of conflict in the housing market is actually slightly lower than This is precisely the manifestation of the acoustic coupling weighting mechanism—different adjacent channels have different noise transmission capabilities and should be treated differently when assessing risks.

[0097] For guest rooms without assigned neighbors ( The system directly commands This indicates no risk of conflict. For example, if the hotel... All guest rooms on the floors above are currently vacant. None of the house's neighbors were assigned a place. , .

[0098] After the adjacent conflict risk index is calculated, the system maps the index to an adjacent compatibility correction coefficient. When assigned neighbors exist, a linear mapping function is used to inversely map the neighbor conflict risk index to a neighbor compatibility correction coefficient, so that the higher the neighbor conflict risk index, the lower the neighbor compatibility correction coefficient. Neighbor compatibility correction coefficient. The specific calculation formula is as follows:

[0099] ; in This is an index representing the risk of conflict between neighboring areas. To correct the upper bound of the coefficient, This is the lower bound of the correction factor, which is greater than zero, and the base value for which no correction is applied is... In this embodiment, , After substituting, it is simplified to (when When there are no assigned neighbors ( When ), set directly That is, the adjacent compatibility correction coefficient takes the baseline value without correction, keeping the original matching score unchanged.

[0100] The design of this linear mapping function has the following technical implications. When (When there are already assigned neighbors, but all neighbors' schedules are fully compatible with the current traveler) This indicates that the room was awarded because the schedules of the guests in the adjacent rooms were compatible. The system should increase the weight of the room as a reward. This increase reflects that, assuming a zero risk of a "quietness mutation," the room not only meets the traveler's own accommodation preferences but also offers a more reliable stay due to its compatibility with the surrounding neighbors' schedules; therefore, the system should recommend the room more actively. (When there are already assigned neighbors and all neighbors' schedules severely conflict with the current traveler's) This indicates that the guest room has been significantly downgraded to its original score due to an extremely high risk of adjacent conflicts. For example, if a single player's original composite matching score is as high as... The risk index of adjacent conflicts for each guest room is: The corrected score will then drop to If a guest room scores significantly lower than other guest rooms with lower conflict risks, the system will no longer prioritize recommending that guest room, thus preventing adjacent noise conflicts from the source of allocation.

[0101] Following the example above, Neighborhood conflict risk index Substitute the values ​​into the formula to calculate its adjacent compatibility correction coefficient:

[0102] ; The correction factor is less than ,show The property faces a higher risk of a quietness mutation due to the presence of assigned neighbors who are active at night and have families with children; its original match score will be compressed to approximately [missing value]. .

[0103] Neighborhood conflict risk index Substitute into the formula:

[0104] ; The correction factor for the house is Slightly higher The house This reflects that its risk of neighboring conflicts is slightly lower than house.

[0105] If there is a guest room without an assigned neighbor (e.g.) (room), then The original score remains unchanged.

[0106] when When it is in the middle value, exist Linear interpolation within the interval enables a continuously differentiable mapping of conflict risk. This means that the impact of adjacent conflict risk on recommendation ranking is not a binary "yes or no" decision, but rather changes continuously with the degree of conflict; the higher the conflict risk, the greater the weight reduction; the lower the conflict risk, the greater the weight increase. This continuous mapping design allows the system to make smooth and reasonable recommendation adjustments under various conflict scenarios, avoiding abrupt changes in recommendation results that might be caused by hard threshold judgments.

[0107] In actual hotel operations, the range of values ​​for the adjacent compatibility correction coefficient is as follows: Labeling can be achieved through statistical analysis of historical hotel complaint data. For example, hotels can extract noise complaint records from the PMS system for the past year, perform correlation analysis between the allocation of complaints at the time of occurrence (which guests were assigned to which adjacent rooms) and the complaint type, and statistically analyze the complaint incidence rate under different conflict combinations, thereby inferring a reasonable lower bound for de-weighting and an upper bound for weighting. and This is the default recommended value; hotels can adjust it according to their own customer base and building sound insulation level.

[0108] The score correction module is used to multiply the original comprehensive matching score using the adjacent compatibility correction coefficient to obtain the corrected comprehensive matching score. In the above process, the system has generated two independent sets of intermediate results: one set is the original comprehensive matching score calculated for each candidate room based on the guest's accommodation intention and dynamic environmental profile; the other set is the adjacency compatibility correction coefficient calculated for each candidate room based on the daily routines and acoustic coupling strength of guests already assigned to adjacent rooms. These two sets of results reflect information from two different dimensions: the degree of matching between the candidate room's own environment and the guest's preferences, and the noise coupling risk brought by guests already assigned to the surrounding area. The current task is to multiply and fuse these two independently generated intermediate results so that the final room recommendation ranking can simultaneously take into account the guest's personal preference satisfaction and the daily routine compatibility of adjacent rooms.

[0109] In its implementation, the system assigns each available guest room... , its original comprehensive matching score Adjacent compatibility correction factor Multiply to calculate the corrected overall matching score. :

[0110] ; in This is the original comprehensive matching score, with a value range of [value range missing]. ; This is the adjacent compatibility correction coefficient, with a value range of [value range missing]. .therefore, The theoretical range of values ​​is .

[0111] The technical significance of this multiplication correction lies in: the correction coefficient. As a multiplicative factor applied to the original matching score, it applies directional and conditional adjustments to the recommendation priority of guest rooms. When the risk of adjacent conflicts among guest rooms is low ( ), correction factor greater than The original score is amplified, and the system is more inclined to recommend that room; when the risk of adjacent conflicts between rooms is high ( The correction factor is less than The original score is compressed, and the system lowers the priority of recommending that room. The direction and magnitude of the correction are determined in real time by the daily routines of the rooms already assigned on the adjacency graph, rather than by fixed rules or static parameters preset in advance.

[0112] Using the aforementioned hotel example, travelers (Business travel, probability of being an early riser) The revision process for each candidate room is as follows.

[0113] Original overall matching score of the house Points, Adjacency Compatibility Correction Coefficient Substitute the values ​​into the formula to calculate the corrected overall matching score:

[0114] ; As can be seen, although Fang scored the highest in the original match ( (points), because of its left neighbor The room has been allocated to nighttime active group travelers, and the neighbor above... Although the apartment is designed for early risers and early bedtimes, it is only connected by a shared floor slab, and the neighbor on the right... The room is already occupied by families with children, and the overall conflict risk index has reached [number missing]. This results in the correction factor being only [missing information]. The original score was compressed to approximately [percentage missing]. The corrected score dropped to This significant deweighting reflects the system's thorough assessment of the potential risk of sudden changes in quietness after occupancy of the room, especially for early-to-bed, early-to-rise business travelers. Assigned to nighttime active passengers Next door, passengers Noise from activities late at night (volume may reach) to Through shared walls (sound insulation) to ) Behind the room, the traveler The actual ambient noise will be determined from the current sensor readings. Rising sharply to to Seriously deviating from passenger safety standards Target quietness .

[0115] Original overall matching score of the house Points, Adjacency Compatibility Correction Coefficient Substitute into the formula:

[0116] ; The corrected score for the house is Divide, and The house The scores are very close but slightly lower. It should be noted that... The raw score of the room ( (score) lower than house( (points), the difference is Points, but its correction factor ( (Higher than) house( The difference is The difference in the correction coefficient partially compensated for the gap in the original scores, making the two rooms almost equal in the corrected ranking. In practical applications, if... If the neighborhood's lifestyle is more favorable (e.g., all neighbors are early risers), the correction factor may increase to [missing value]. Above, at this time The corrected score for the house will exceed The room, thus reversing the order.

[0117] To more clearly demonstrate the reversal effect of the correction coefficient on the ranking, let's assume there is another candidate room. The house, its original comprehensive matching score However, all the neighbors in this room have been assigned to early-to-bed, early-to-rise travelers, and all are adjacent to each other with shared walls, indicating a high risk of neighborly conflict. (Current Passenger) have The probability of being an early riser is the conflict value with neighbors who are also early risers. Correction factor Substitute into the formula:

[0118] ; The corrected score for this room is as high as Points, far exceeding The house divide and sum The house Points. In the original rankings before correction, House ranked Fanghe After the house ( However, after adjacent compatibility correction, Fang rose to the top spot ( The physical meaning of this reversal of order is very clear: although The room's independent environment is slightly less suitable than Fanghe The room was available, but its location was different from the lifestyles of the surrounding neighbors and travelers. Highly compatible, the risk of sudden changes in quietness after move-in is extremely low; the actual living experience is the most reliable indicator. The system will... The fact that this room has been promoted to the top of the recommendations is a proactive response to the dynamic occupancy coupling effect of adjacent rooms.

[0119] For guest rooms without assigned neighbors House, assuming its original comprehensive matching score Points, correction coefficient The corrected score is... The original score remains unchanged. Because there is a lack of information on the schedules of adjacent rooms, the system neither upgrades nor downgrades this room's ranking, adopting a neutral approach.

[0120] For guest rooms with downgraded data confidence The house, its original comprehensive matching score Points (included) (confidence penalty), correction coefficient (Assuming its adjacent conflict risk is slightly lower) If the room is corrected, the score will be adjusted accordingly. The room's ranking dropped further due to both a confidence penalty and a downgrade caused by adjacent conflicts.

[0121] The revised overall matching scores of all the candidate rooms are summarized, and the ranking results are as follows: house( point) house( point) house( point) house( point) house( (Points). Compared to the previous sorting based solely on independent matches ( house point house point house point house point house Compared to the previous ranking (points), the recommendation order has changed significantly. The one that was originally ranked first... The house dropped to third place due to a higher risk of neighborly conflicts; it was originally ranked third. Rooms with the lowest risk of adjacent conflicts have risen to the top of the rankings. This change in ranking is not a random fluctuation, but rather a systematic reflection of the objective impact of the dynamic occupancy coupling effect of adjacent rooms on the actual living experience of guest rooms.

[0122] A key characteristic of this multiplication correction mechanism is its preservation of the original matching score. When When there are no assigned neighbors or the correction coefficient is exactly the baseline value, the corrected score equals the original score, and the system's behavior degenerates into the traditional independent matching pattern. This means that the difference between this method and the traditional system only becomes apparent in scenarios where assigned neighbors exist, and it does not introduce additional computational biases in cases without adjacency constraints. Meanwhile, The upper realm and the lower realm This ensures that the correction is within a reasonable range—even in the most extreme conflict scenarios, the room score will not be completely reset to zero (the minimum is 1 / 3 of the original score). The guest retains the final right to accept the room; even in the most ideal compatibility scenario, the room score will not be over-amplified (the maximum score is the original score). This avoids the complete monopoly of recommendation results by a few compatible rooms.

[0123] The recommendation output module is used to sort each candidate room according to the corrected comprehensive matching score, and output the final room recommendation result by combining the preset hard constraint check conditions.

[0124] In this module, the system first matches all candidate guest rooms according to the revised overall matching score. Sort the rooms in descending order to generate a candidate room ranking list. Following the previous example, the corrected sorting result is: house( point) house( point) house( point) house( point) house( (Points). This ranking result reflects the consideration of passengers simultaneously. After considering the coupling risks between individual accommodation preferences and dynamic occupancy of adjacent rooms, each guest room has a different impact on the traveler's needs. The overall suitability is considered. At this point, the top-ranked room in the sorted list only indicates "optimal under the current known conditions" and cannot be directly used as the final recommended room output, because the room still needs to meet the hard constraint check conditions at the operational level.

[0125] Subsequently, the system performs hard constraint checks on each room in the candidate room ranking list. These hard constraint checks include room availability constraints, accessibility compliance constraints, and fire evacuation compliance constraints. To facilitate subsequent processing, the system can set a hard constraint exclusion flag for each candidate room. When the guest room passes all hard constraint checks, If any hard constraint is not satisfied, then let Then remove the room from the sort list.

[0126] Room availability constraints are used to reconfirm that guest rooms are still available for allocation at the time of recommendation. Although candidate guest rooms have passed the initial room availability screening in the preceding processes, the hotel front desk, self-check-in terminals, and mobile devices may be working concurrently, and the room status may still change between the completion of the sorting calculation and the final locking. Therefore, the system checks the current room availability flag for each guest room again before outputting the recommendation results. Is it still "assignable"? For example, if... If a room is prioritized for allocation to other guests by another front desk terminal after the sorting is generated, or if it is updated to "cleaning" by the housekeeping department, then the current room status of that room no longer meets the recommended criteria. , Even rooms with the highest scores will be rejected. This secondary verification avoids the problem of "the system recommending available rooms but the front desk being unable to actually deliver them."

[0127] Accessibility compliance constraints are used to protect accessible room resources and prevent general travelers from occupying rooms reserved for travelers with special needs. Specifically, if a hotel currently has a booked accessible guest, and a candidate room is an accessible room, that room cannot be recommended to a general traveler. For example, if... The room is a accessible room, and will be available later that day. There will be one wheelchair user checking in, so even though The revised overall matching score for the house is higher than Fanghe The system still allows This will remove it from the sort list. The technical effect of this constraint is to ensure that the priority of special resources in hotel operations is not overridden by a pure score sort.

[0128] Fire evacuation compliance constraints are used to further filter out rooms unsuitable for current guests under special room type conditions. For rooms located on basements, at the end of long corridors, or where evacuation distances are close to regulatory limits, if the current guest to be assigned is a family traveling with young children, the system must check whether the evacuation distance from that room to the nearest safe exit is within the regulatory limit. If it exceeds the limit, then... And remove them. Using the current example, passengers... The travel purpose is tagged as business trip, which does not fall under the category of family travel with young children; therefore, this restriction applies to the traveler. This constraint is not triggered during the allocation process. However, within the same check-in batch, if the next guest is a family traveler, this constraint will participate in the selection of their candidate rooms.

[0129] After completing the hard constraint check, the system will... The selected guest rooms are removed from the candidate guest room ranking list, generating a simplified ranking list after constraint filtering. Continuing with the example above, if... The property was excluded due to changes in its condition. Rooms are reserved for travelers with special needs due to accessibility compliance restrictions, so the simplified sorting list becomes: house( point) house( point) house( (Points). At this point, the system will output and display the top-ranked room after streamlining as the final recommended room; that is, the final recommended room in the current example is... house.

[0130] When displaying the recommendation results, the system not only outputs the final recommended room number but also simultaneously provides a summary of the recommendation reasons to improve the comprehensibility of the results for front desk staff and guests. A summary of reasons for recommending a property may include: quietness score. Score, original comprehensive matching score Points, Adjacency Compatibility Correction Coefficient Corrected overall matching score The system assigns points and explains that the room's lower priority is due to the presence of active nighttime travelers in the left-hand neighboring area and families with children in the right-hand neighboring area. This allows front desk staff to clearly see the specific reasons why "the room itself is quiet, but the surrounding adjacent areas have a higher risk of occupancy." If the guest accepts this recommendation, the system continues to lock the room; if the guest refuses, for example, by insisting on staying on a higher floor or making new special requests, the system will remove the room from the priority list. Select the next room as an alternative recommendation, for example, from The room is then listed as the second recommendation, and its matching score and adjacent compatibility information are displayed again until the traveler accepts or the candidate list is exhausted.

[0131] After receiving confirmation from the traveler, the system updates the room status of the final recommended room to "assigned" in the hotel property management system to complete the room lock. In the specific PMS implementation, the "assigned" status can be further subdivided into "assigned but not yet checked in" to distinguish it from the "checked in" status. Using the current example, if the traveler... accept The system will Update to the assigned status and write the room lock timestamp and guest number to prevent other terminals from reassigning the same room.

[0132] After the room status is locked, the system also writes the guest's activity level probability distribution into the allocated record cache, which will serve as guest activity level information for subsequent concurrent guest check-in analysis. Using the current example, the system will... probability distribution of daily routine types Write to the allocated record cache. To improve subsequent read efficiency, the system can also simultaneously cache the specific work-rest type label corresponding to this probability distribution. This refers to the early-to-bed, early-to-rise type. This way, when subsequent passengers... Enter the allocation process and evaluate house, House or Housing and other properties When there are adjacent guest rooms, the system can directly connect the guests. Considered The assigned neighbors of the house are used to participate in the calculation of a new neighborhood conflict risk index based on their daily routine information.

[0133] Once the current passenger is assigned a room, the system returns to the ordered processing queue to continue processing the next passenger awaiting assignment, until all passengers in the current concurrent check-in batch have had their rooms recommended and locked. Because room status data and assigned records are updated immediately after each assignment, the calculation of adjacent room conflict risks for subsequent passengers is always based on the latest assignment status, achieving global dynamic linkage in concurrent check-in scenarios.

[0134] In summary, by sorting each candidate room according to the corrected comprehensive matching score, overlaying room availability constraints, accessibility compliance constraints, and fire evacuation compliance constraints, outputting the final recommendation result, and performing room status locking and cache write-back, the system organically combines individual guest preference matching, dynamic occupancy coupling risk control of adjacent rooms, and actual hotel operational constraints. This not only avoids the problem of sudden changes in quietness such as "recommending a quiet room but it is not quiet after check-in," but also reduces the room change rate caused by adjacent noise conflicts, improves the effective utilization rate of room resources, and enhances the trust of front desk staff and guests in the intelligent allocation results, thus fully realizing a hotel check-in management method based on intelligent interaction.

[0135] Example 2:

[0136] like Figure 3 As shown, a hotel check-in management method based on intelligent interaction includes the following steps: Obtain accommodation intention data and travel purpose tags of passengers currently awaiting allocation, as well as dynamic environmental profiles of each candidate room and information on the daily routines of guests in already allocated rooms; Based on accommodation intent data and dynamic environmental profiles, multi-dimensional independent matching is performed to calculate the original comprehensive matching score for each candidate room. Obtain a pre-constructed guest room adjacency graph, which uses hotel guest rooms as vertices and physical adjacency relationships as edges, with edge weights representing the acoustic coupling strength between adjacent guest rooms; Infer the probability distribution of the daily routine type of the current passengers to be assigned based on the travel purpose label; For each candidate room, its assigned neighbors in the room adjacency graph are extracted. If there are assigned neighbors, the expected conflict level is calculated based on the guest activity status information and activity type probability distribution of the assigned neighbors. The expected conflict level is then weighted by the edge weight between the assigned neighbors and the candidate room to obtain the adjacency conflict risk index. The adjacency conflict risk index is then mapped to the adjacency compatibility correction coefficient. The original comprehensive matching score is corrected using the adjacency compatibility correction coefficient to obtain the corrected comprehensive matching score; The candidate guest rooms are sorted according to their comprehensive matching scores, and the final guest room recommendation results are output based on the preset hard constraint check conditions.

[0137] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0138] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0139] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A hotel check-in management system based on intelligent interaction, characterized in that, include: The data acquisition module is used to acquire the accommodation intention data and travel purpose tags of the passengers currently to be assigned, as well as the dynamic environmental profiles of each candidate room and the rest and activity status information of the guests in the assigned rooms; The multi-dimensional matching module is used to perform multi-dimensional independent matching based on accommodation intent data and dynamic environmental profiles, and calculate the original comprehensive matching score of each candidate room. The relationship graph acquisition module is used to acquire a pre-built guest room adjacency relationship graph. The guest room adjacency relationship graph has hotel guest rooms as vertices and physical adjacency relationships as edges, and the edge weights represent the acoustic coupling strength between adjacent guest rooms. The daily routine inference module is used to infer the probability distribution of the daily routine type of the current passenger to be assigned based on the travel destination label; The compatibility analysis module is used to extract the assigned neighbors of each candidate room in the room adjacency graph. If there are assigned neighbors, the expected conflict level is calculated based on the guest activity status information and activity type probability distribution of the assigned neighbors. The expected conflict level is weighted by combining the edge weight between the assigned neighbors and the candidate room to obtain the adjacency conflict risk index. The adjacency conflict risk index is then mapped to the adjacency compatibility correction coefficient. The score correction module is used to correct the original comprehensive matching score using the adjacency compatibility correction coefficient, so as to obtain the corrected comprehensive matching score. The recommendation output module is used to sort each candidate room according to the corrected comprehensive matching score, and output the final room recommendation result by combining the preset hard constraint check conditions.

2. A hotel check-in management system based on intelligent interaction according to claim 1, characterized in that: Obtain a pre-built adjacency graph of guest rooms, including: Based on the hotel building topology data, all hotel rooms are taken as vertices, and edges are established between the vertices of the rooms that meet the physical adjacent conditions of sharing walls, sharing floors, or diagonally offset, to construct an undirected weighted graph as the adjacency graph of the rooms. Each edge is assigned a corresponding acoustic coupling weight based on its adjacency type. The acoustic coupling weight is largest for edges that share a wall and smallest for edges that are diagonally offset.

3. A hotel check-in management system based on intelligent interaction according to claim 1, characterized in that: Based on accommodation intent data and dynamic environmental profiles, multi-dimensional independent matching is performed to calculate the original comprehensive matching score for each candidate room, including: Semantic entity recognition is performed on accommodation intent data to extract preference intensity coefficients under each matching dimension and normalize them to generate traveler intent preference weight vectors. At the same time, accommodation intent data is mapped to target parameter vectors for each matching dimension. Obtain static attributes and real-time sensor data for each candidate guest room, and construct a dynamic environmental profile accordingly; For each matching dimension, calculate the single-dimensional matching score between the current state value in the dynamic environment profile and the target parameter of the corresponding dimension; The original comprehensive matching score is obtained by weighting and summing the single-dimensional matching scores of each dimension using the passenger intent preference weight vector.

4. A hotel check-in management system based on intelligent interaction according to claim 1, characterized in that: Based on the travel destination label, the probability distribution of the current traveler's daily routine type is inferred, including: Obtain the prior transition probability matrix from the preset travel purpose to the rest type. The rows of the prior transition probability matrix correspond to different travel purpose categories, and the columns correspond to different rest types. Using the travel destination label of the passenger to be assigned as an index, the corresponding row is extracted from the prior transition probability matrix and used as the activity type probability distribution vector of the passenger to be assigned. Each component in the activity type probability distribution vector represents the probability that the passenger belongs to the corresponding activity type.

5. A hotel check-in management system based on intelligent interaction according to claim 1, characterized in that: The expected degree of conflict is calculated based on the resident activity status information and activity type probability distribution of the assigned neighbors, including: Obtain a preset work-rest conflict matrix, where the elements of the work-rest conflict matrix represent the degree of conflict between different work-rest types; Extract the specific daily routine type tag of the assigned neighbor from the resident's daily routine status information; Using the probability distribution vector of work and rest types as weights, the elements in the work and rest conflict matrix corresponding to the column of the defined work and rest type label are weighted and summed to obtain the expected degree of conflict.

6. A hotel check-in management system based on intelligent interaction according to claim 5, characterized in that: By weighting the expected conflict level by combining the edge weights between the assigned neighbors and the candidate room, an adjacency conflict risk index is obtained, which includes: Extract the acoustic coupling weights of all assigned neighbors of the candidate guest room; The neighbor conflict risk index is calculated by weighting the expected conflict levels of each assigned neighbor according to the corresponding acoustic coupling weights.

7. A hotel check-in management system based on intelligent interaction according to claim 1, characterized in that: Mapping the adjacent conflict risk index to an adjacent compatibility correction coefficient includes: When there are already assigned neighbors, a linear mapping function is used to reverse the neighbor conflict risk index to a neighbor compatibility correction coefficient, so that the higher the neighbor conflict risk index, the lower the neighbor compatibility correction coefficient.

8. A hotel check-in management system based on intelligent interaction according to claim 3, characterized in that: When constructing a dynamic environmental profile based on the static attributes of each candidate guest room and real-time sensor data, it also includes: Check for missing or faulty real-time sensor data for each candidate guest room; If it exists, the average value of the group of rooms on the same floor and facing the same direction as the candidate room during the same historical period is used as the replacement value, and the data confidence is downgraded in the dynamic environmental profile of the candidate room. When calculating the original comprehensive matching score, for candidate rooms marked with data confidence downgrade, a preset confidence penalty factor is multiplied by the weighted summation result to reduce their original comprehensive matching score.

9. A hotel check-in management system based on intelligent interaction according to claim 1, characterized in that: The candidate rooms are sorted according to the revised comprehensive matching score, and combined with the preset hard constraint check conditions, the final room recommendation results are output, including: The candidate rooms are sorted in descending order according to the revised comprehensive matching score to generate a candidate room sorting list; Hard constraint checks are performed on the guest rooms in the candidate guest room ranking list in turn. The hard constraint check conditions include room availability constraints, accessibility compliance constraints, and fire evacuation compliance constraints. Candidate rooms that fail the hard constraint check are removed, and the top-ranked room is displayed as the final recommended room. After receiving the passenger's confirmation instruction, the room status of the final recommended room is updated to the assigned status, and the probability distribution of the passenger's daily routine type is written into the assigned record cache as guest daily routine status information for subsequent concurrent guest check-in analysis.

10. A hotel check-in management method based on intelligent interaction, characterized in that, Includes the following steps: Obtain accommodation intention data and travel purpose tags of passengers currently awaiting allocation, as well as dynamic environmental profiles of each candidate room and information on the daily routines of guests in already allocated rooms; Based on accommodation intent data and dynamic environmental profiles, multi-dimensional independent matching is performed to calculate the original comprehensive matching score for each candidate room. Obtain a pre-constructed guest room adjacency graph, which uses hotel guest rooms as vertices and physical adjacency relationships as edges, with edge weights representing the acoustic coupling strength between adjacent guest rooms; Infer the probability distribution of the daily routine type of the current passengers to be assigned based on the travel purpose label; For each candidate room, its assigned neighbors in the room adjacency graph are extracted. If there are assigned neighbors, the expected conflict level is calculated based on the guest activity status information and activity type probability distribution of the assigned neighbors. The expected conflict level is then weighted by the edge weight between the assigned neighbors and the candidate room to obtain the adjacency conflict risk index. The adjacency conflict risk index is then mapped to the adjacency compatibility correction coefficient. The original comprehensive matching score is corrected using the adjacency compatibility correction coefficient to obtain the corrected comprehensive matching score; The candidate guest rooms are sorted according to their comprehensive matching scores, and the final guest room recommendation results are output based on the preset hard constraint check conditions.

Images