A cloud computer recommendation method, system, device, storage medium and product

Abstract

The application discloses a cloud computer recommendation method, system, device, storage medium and product, comprising the following steps: according to historical behavior records of users on cloud computers, a collaborative filtering algorithm is used to calculate a collaborative filtering score of each user on each cloud computer; user features are extracted, and users are clustered according to the user features to obtain a plurality of user groups; cloud computer features are extracted, and a content-based recommendation algorithm is used to calculate a content similarity score of each user on each cloud computer according to the cloud computer features; a clustering feedback score of each user on each cloud computer is obtained according to a feedback score of each user group on the configuration of each cloud computer; and a recommendation score of each user on each cloud computer is calculated according to the collaborative filtering score, the content similarity score and the clustering feedback score. The embodiment of the application can ensure that the recommendation result meets the personalized needs of users and is verified by a group, thereby improving the user availability of the recommendation result.

Description

Technical Field

[0001] This invention relates to the field of intelligent recommendation technology, and in particular to a cloud computer recommendation method, system, device, storage medium, and product. Background Technology

[0002] In cloud computing resource usage scenarios, user interaction with resources is relatively limited, and the user base may not be as large as that of consumer products. This can lead to an extremely sparse user-resource interaction matrix, making it difficult for algorithms to find enough similar users or resources for effective recommendations.

[0003] In addition, the demand for cloud resources changes over time and may be affected by various factors such as business cycles, promotional activities, and technology updates. Current recommendation algorithms often focus on static historical behavior analysis and have limited adaptability to real-time changes. Summary of the Invention

[0004] The embodiments of the present invention aim to provide a cloud computer recommendation method, system, device, storage medium and product that can ensure that the recommendation results not only meet the user's personalized needs, but are also verified by the group, thereby improving the user usability of the recommendation results.

[0005] In a first aspect, embodiments of the present invention provide a cloud computer recommendation method, including:

[0006] Based on the user's historical behavior records of the cloud computers, a collaborative filtering algorithm is used to calculate the collaborative filtering score for each user on each of the cloud computers.

[0007] Extract user features, and cluster users based on these features to obtain several user groups;

[0008] Extract cloud computer features, and based on the cloud computer features, use a content-based recommendation algorithm to calculate the content similarity score for each user to each cloud computer;

[0009] Based on the feedback scores of each user group on the configuration of each cloud computer, the clustering feedback score of each user on each cloud computer is obtained.

[0010] Based on the collaborative filtering score, the content similarity score, and the clustering feedback score, a recommendation score for each user for each of the cloud computers is calculated.

[0011] As an improvement to the above solution, the step of calculating the collaborative filtering score for each user on each cloud computer using a collaborative filtering algorithm based on the user's historical behavior records of the cloud computers includes:

[0012] Based on users' historical behavior records of cloud computers, calculate the first similarity between users and the second similarity between cloud computers;

[0013] Based on the first similarity, the second similarity, and the user's rating behavior for historical interactive cloud computers, the predicted rating of the user for all cloud computers is calculated, and the collaborative filtering score of each user for each cloud computer is obtained.

[0014] As an improvement to the above scheme, the step of extracting user features and clustering users based on these features to obtain several user groups includes:

[0015] Extract user characteristics;

[0016] Based on the user characteristics, the ant colony algorithm is used to obtain the initial cluster centers of the users; wherein, the mutation factor of the ant colony algorithm is defined using the golden sine algorithm.

[0017] Based on the initial cluster centers and the user characteristics, the K-means clustering algorithm is used to cluster the users, resulting in several user groups.

[0018] As an improvement to the above solution, the step of extracting cloud computer features and, based on these features, employing a content-based recommendation algorithm to calculate the content similarity score for each user to each of the cloud computers includes:

[0019] Extracting features from cloud computers;

[0020] Based on the cloud computer characteristics described in the user's historical interactions with the cloud computer, construct a user interest vector;

[0021] Calculate the third similarity between the cloud computer features of each cloud computer and the user interest vector to obtain the content similarity score of each user for each cloud computer.

[0022] As an improvement to the above scheme, the step of obtaining the clustering feedback score of each user for each cloud computer based on the feedback score of each user group for the configuration of each cloud computer includes:

[0023] Obtain the first feedback score of each user in each user group on each of the cloud computers on the preset configuration items;

[0024] Based on the first feedback score, calculate the average feedback score for each user on all configuration items;

[0025] Based on the preset user weights and the average feedback score, the clustering feedback score of each user group for the configuration of each cloud computer is calculated.

[0026] As an improvement to the above scheme, the step of calculating the recommendation score for each user for each of the cloud computers based on the collaborative filtering score, the content similarity score, and the clustering feedback score includes:

[0027] Based on preset algorithm weights, the collaborative filtering score, the content similarity score, and the clustering feedback score are weighted and fused to obtain the recommendation score for each user for each cloud computer.

[0028] Based on the user's recommendation scores for all cloud computers, the N cloud computers with the highest recommendation scores are selected to generate cloud computer recommendation results for the user.

[0029] Secondly, embodiments of the present invention provide a cloud computer recommendation system, including:

[0030] The collaborative filtering algorithm application module is used to calculate the collaborative filtering score for each user on each cloud computer based on the user's historical behavior records on the cloud computers using a collaborative filtering algorithm.

[0031] The user clustering module is used to extract user features and cluster users according to the user features to obtain several user groups;

[0032] The content recommendation algorithm application module is used to extract cloud computer features and, based on the cloud computer features, employ a content-based recommendation algorithm to calculate the content similarity score for each user to each of the cloud computers.

[0033] The clustering feedback application module is used to obtain the clustering feedback score of each user for each cloud computer based on the feedback score of each user group for the configuration of each cloud computer.

[0034] The algorithm fusion module is used to calculate the recommendation score for each user for each of the cloud computers based on the collaborative filtering score, the content similarity score, and the clustering feedback score.

[0035] Thirdly, embodiments of the present invention provide a cloud computer recommendation device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the cloud computer recommendation method as described above.

[0036] Fourthly, embodiments of the present invention provide a computer-readable storage medium comprising a stored computer program, wherein the computer program, when running, controls the device where the computer-readable storage medium is located to execute the cloud computer recommendation method as described above.

[0037] Fifthly, embodiments of the present invention provide a computer program product, the computer program product including a computer program or computer instructions, wherein when the computer program or computer instructions are executed by a processor, the cloud computer recommendation method described above is performed.

[0038] Compared with existing technologies, the cloud PC recommendation method, system, device, storage medium, and product disclosed in this invention calculate the collaborative filtering score for each user on each cloud PC using a collaborative filtering algorithm based on the user's historical behavior records of cloud PCs; extract user features and cluster users according to these features to obtain several user groups; extract cloud PC features and calculate the content similarity score for each user on each cloud PC using a content-based recommendation algorithm based on these features; obtain the clustering feedback score for each user on each cloud PC based on the feedback scores of each user group on the configuration of each cloud PC; and calculate the recommendation score for each user on each cloud PC based on the collaborative filtering score, the content similarity score, and the clustering feedback score. Using the embodiments of this invention, it is possible to ensure that the recommendation results not only meet the user's personalized needs but are also validated by the group, thereby improving the user usability of the recommendation results. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the steps of a cloud computer recommendation method provided in an embodiment of the present invention;

[0040] Figure 2 This is a schematic diagram of the structure of a cloud computer recommendation system provided in an embodiment of the present invention;

[0041] Figure 3 This is a schematic diagram of the structure of a cloud computer recommendation device provided in an embodiment of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0043] In the description and claims, it should be understood that the terms "first," "second," etc., used in the description and claims are only for the purpose of distinguishing the description of the same technical features, and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated, nor necessarily the order of description or chronological order. The terms are interchangeable where appropriate. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature.

[0044] This invention provides a method for recommending cloud computers. Please refer to [link / reference]. Figure 1 In this embodiment, the cloud computer recommendation method is specifically executed through steps S1 to S5:

[0045] S1. Based on the user's historical behavior records of the cloud computers, a collaborative filtering algorithm is used to calculate the collaborative filtering score of each user for each of the cloud computers.

[0046] S2. Extract user features, and cluster users according to the user features to obtain several user groups;

[0047] S3. Extract cloud computer features, and based on the cloud computer features, use a content-based recommendation algorithm to calculate the content similarity score of each user to each cloud computer.

[0048] S4. Based on the feedback scores of each user group on the configuration of each cloud computer, obtain the clustering feedback score of each user on each cloud computer.

[0049] S5. Calculate the recommendation score for each user for each of the cloud computers based on the collaborative filtering score, the content similarity score, and the clustering feedback score.

[0050] In some preferred embodiments, data is automatically collected from various sources through deployed distributed log collection systems (such as Fluentd and Logstash) and cloud platform API interfaces to obtain historical behavior records and basic data that can be used to extract user characteristics and cloud computer characteristics, ensuring the timeliness and integrity of the data. Simultaneously, real-time data transmission and preliminary filtering are achieved through event stream processing frameworks (such as Apache Kafka and Amazon Kinesis).

[0051] Collaborative filtering (CF) primarily uses users' historical behavior and preferences to analyze the similarities between users or between cloud computers to predict users' preferences for non-interactive cloud computers and generate personalized recommendations.

[0052] The collaborative filtering score calculated by the collaborative filtering algorithm can capture the similarity of behavioral patterns among users and uncover potential interests. This uncovering process is not limited to historical cloud computers, but can also find implicit associations with other users or different cloud computers.

[0053] In some preferred embodiments, user features are extracted through user behavior models. User behavior models, which can extract user features through time series analysis or user profiling, can characterize user habits, preferences, and behavioral patterns. This includes not only explicit behavioral pattern recognition but also implicit prediction of user needs.

[0054] Time series analysis is a key tool for understanding how user behavior evolves over time. This process is particularly important in cloud computing resource recommendation systems, as it reveals users' resource usage habits and preferences across different time periods, providing a basis for dynamic resource allocation and optimization.

[0055] For example, time series forecasting models using ARIMA (Autoregressive Integral Moving Average) or seasonal decomposition can identify daily, weekly, and annual cyclical patterns in user usage. For instance, many users may have significant usage peaks in the mornings and afternoons of weekdays, while exhibiting different usage rhythms on weekends.

[0056] By using seasonal decomposition methods, such as the X-13ARIMA-SEATS model, seasonal factors, such as changes in usage patterns that may occur during holidays and school breaks, can be extracted separately. This is crucial for proactively adjusting resource allocation strategies.

[0057] In this embodiment of the invention, to further refine the user group, a clustering algorithm is used to group users with similar characteristics into the same group, with each user group representing a typical behavioral pattern. By grouping user behavior data, user groups with similar usage habits and needs can be identified.

[0058] By dividing a large number of users into user groups, this embodiment of the invention can effectively improve the coverage of cold-start users in the recommendation process and provide a foundation for subsequent aggregation of cloud computer configuration preferences.

[0059] Preferably, the cloud computer features include key attributes such as hardware configuration (e.g., CPU performance, memory size, storage capacity), price, and average user ratings. In a preferred embodiment of the present invention, if text descriptions are involved in the feature extraction process, methods such as TF-IDF and word embedding (e.g., Word2Vec, GloVe) are used to convert the text into vectors.

[0060] Content-based filtering (CBF) relies on the characteristics of items themselves, such as the hardware configuration, price, and user reviews of cloud computers, to recommend new items with similar characteristics to items that a user has historically liked. Key steps include feature extraction (such as TF-IDF and word vectors) and similarity calculation (such as cosine similarity), and then recommending highly similar items based on a user interest model (composed of features of items the user has historically liked).

[0061] By calculating content similarity scores based on content, we can ensure that the recommended cloud computers meet the user's professional needs in terms of overall features, thus achieving accurate matching.

[0062] In some preferred embodiments, an instant feedback collection platform is also provided in the cloud service, on which users can provide feedback on their experience using the cloud computer, including but not limited to performance, ease of use, and satisfaction with resource allocation.

[0063] More preferably, the instant feedback collection platform integrates an online chatbot or customer service representative so that users can get immediate help when they encounter problems. It also provides real-time communication features, such as screen sharing, to facilitate remote diagnosis and problem-solving by technical personnel.

[0064] By collecting these feedback scores in real time, service providers can quickly respond to user needs, optimize cloud computing resource allocation, and provide users with a more personalized and efficient cloud service experience. Furthermore, considering the potential for individual behavior to be accidental, clustering feedback scores can smooth out recommendation biases caused by abnormal individual behavior.

[0065] Collaborative filtering scores are used for recommendations based on user or cloud computer similarity, content similarity scores are used for recommendations based on individual preference configurations, and clustering feedback scores are used for recommendations based on group consensus. The final recommendation score, obtained by combining these three scores, can dynamically adjust and recommend the most suitable cloud resource configuration based on the user's usage patterns, historical feedback, and the characteristics of their group, achieving a balance between personalization and universality.

[0066] The above scheme, by simultaneously considering user behavior patterns, personal preferences, and group consensus, can verify the rationality of recommendations from multiple perspectives, avoid the bias that may arise from a single algorithm, and ensure that the recommendation results not only meet the user's personalized needs but are also validated by the group, thereby improving the user usability of the recommendation results.

[0067] As a preferred implementation, step S1 involves calculating the collaborative filtering score for each user on each cloud computer based on the user's historical behavior records of the cloud computers using a collaborative filtering algorithm, including:

[0068] Based on users' historical behavior records of cloud computers, calculate the first similarity between users and the second similarity between cloud computers;

[0069] Based on the first similarity, the second similarity, and the user's rating behavior for historical interactive cloud computers, the predicted rating of the user for all cloud computers is calculated, and the collaborative filtering score of each user for each cloud computer is obtained.

[0070] In this embodiment of the invention, the collaborative filtering algorithm can be implemented through two paths: user-to-user collaborative filtering and cloud computer-to-cloud computer collaborative filtering.

[0071] The principle of user-to-user collaborative filtering is to find other users with similar interests to the target user, and then recommend cloud computers that similar users like but the target user has not yet interacted with.

[0072] In some preferred embodiments, the user-user collaborative filtering predicts the first collaborative filtering score of user u for cloud computer i. Represented as:

[0073] (1);

[0074] in, The set of users whose first similarity to user u is greater than a preset first threshold; The first similarity between user u and user v; User v's rating of cloud computer i; To The weighted normalization factor is used to ensure that the predicted score is within a reasonable range.

[0075] The principle of cloud PC collaborative filtering is based on the user's rating or interaction behavior with cloud PCs, calculating the similarity between cloud PCs, and then recommending other cloud PCs similar to those with the user's historical interactions.

[0076] Preferably, the second collaborative filtering score predicted by cloud computer-cloud computer collaborative filtering for user u against cloud computer i. Represented as:

[0077] (2);

[0078] in, A collection of user u's historical interactions with cloud computers; The second similarity between cloud computer i and cloud computer j; The rating of the historical interactive cloud computer by user u; This is a normalization term used to ensure the comparability of predicted scores.

[0079] In some preferred embodiments, the first collaborative filtering score or the second collaborative filtering score is used as the collaborative filtering score for the user on each cloud computer. In other preferred embodiments, the first collaborative filtering score or the second collaborative filtering score is weighted and summed to obtain the collaborative filtering score.

[0080] The above scheme introduces both primary similarity between users and secondary similarity between cloud computers, adding a double constraint to rating prediction and making the prediction results more robust and accurate. Furthermore, the bidirectional similarity network allows information to spread and diffuse simultaneously along both user and cloud computer dimensions. Previously isolated users or cloud computers can be integrated into the entire recommendation network through indirect similarity connections, thereby reducing recommendation blind spots.

[0081] In a preferred implementation, step S2 involves extracting user features and clustering users based on these features to obtain several user groups, including:

[0082] Extract user characteristics;

[0083] Based on the user characteristics, the ant colony algorithm is used to obtain the initial cluster centers of the users; wherein, the mutation factor of the ant colony algorithm is defined using the golden sine algorithm.

[0084] Based on the initial cluster centers and the user characteristics, the K-means clustering algorithm is used to cluster the users, resulting in several user groups.

[0085] It should be noted that, in this embodiment of the invention, the ant colony algorithm adopts a mutation strategy based on the Golden Sine Algorithm (GSA).

[0086] In some preferred embodiments, the variation factor is defined using the golden ratio and a sine function, and the variation factor is expressed as:

[0087] (3);

[0088] in, Parameters used to control the intensity of variation; It is a random number.

[0089] In ant colony optimization, the sine function from the golden sine algorithm is first used to randomly initialize the ant positions or paths, thus initializing the pheromone matrix. Then, during ant movement, the pheromone matrix is ​​initialized according to a preset mutation probability. Control the frequency of mutation operations and use the GSA position update formula. The position of each ant is adjusted to ensure that the mutated position is within the solution space.

[0090] Preferably, the number of clusters can be determined by methods such as the elbow rule or the profile coefficient.

[0091] After applying the GSA mutation strategy, the pheromone update rule is adjusted based on the quality of the new solution to reflect its contribution to the solution space. Combining the search characteristics of GSA, the use of heuristic information is adjusted to guide the ants to move to more promising regions. Through multiple iterations, the ants (representing the decision-making process) move between data points and potential cluster centers, update the pheromone concentration, and select the next cluster center. After each iteration, the global optimal solution is updated based on the choices of all ants until a preset number of iterations is reached or the pheromone update converges, ultimately yielding the initial cluster centers.

[0092] Once a set of potential cluster centers is found using the ant colony algorithm, the K-means algorithm can be used for clustering.

[0093] In K-means clustering, each data point is first assigned to the nearest cluster center. Then, the average value of each cluster is recalculated as the new cluster center. The above steps are repeated until the cluster centers no longer change significantly or the maximum number of iterations is reached. This completes the clustering of users and yields user groups.

[0094] The K-means algorithm is highly sensitive to the selection of initial cluster centers, while the ant colony algorithm's exploration and utilization mechanisms can help find more reasonable initial cluster centers. By searching for optimal paths or candidate cluster centers through ant colonies, K-means can be provided with a better starting point, reducing the risk of getting trapped in local optima. K-means may perform poorly when dealing with non-convex datasets or noisy data, while the distributed search and pheromone mechanism of the ant colony algorithm can better explore the data space, helping to identify potential clustering structures in complex data distributions.

[0095] By applying these clustering algorithms, several core user behavior patterns can be extracted from massive amounts of user data, such as the "night owl" user group (preferring to use high-performance resources at night) and the "regular office worker" user group (high usage rate during weekdays). These classifications not only deepen the understanding of user behavior but also provide a basis for customized resource recommendation strategies, ensuring that resources can be more accurately matched to the needs of different user groups.

[0096] The above scheme combines two algorithms, leveraging the global optimization capability of ant colony optimization and the simplicity and speed of K-means to jointly optimize clustering results. Ant colony optimization, by simulating the mechanism of ants searching for food, continuously optimizes paths (or cluster boundaries), resulting in more stable and accurate clustering results.

[0097] As a preferred implementation, step S3 involves extracting cloud computer features and, based on these features, employing a content-based recommendation algorithm to calculate the content similarity score for each user to each of the cloud computers, including:

[0098] Extracting features from cloud computers;

[0099] Based on the cloud computer characteristics described in the user's historical interactions with the cloud computer, construct a user interest vector;

[0100] Calculate the third similarity between the cloud computer features of each cloud computer and the user interest vector to obtain the content similarity score of each user for each cloud computer.

[0101] In this embodiment of the invention, the key steps of the content-based recommendation algorithm include feature extraction (such as TF-IDF, word vectors) and similarity calculation (such as cosine similarity), and then recommending items with high similarity based on the user interest model (composed of features of items that the user has historically preferred).

[0102] For any user, a user interest vector can be constructed based on the characteristics of their historical cloud computer interactions. This is typically a weighted average of the feature vectors of cloud computers that users have previously liked or rated highly. It should be noted that the weights mentioned here can be user ratings or the intensity of interaction.

[0103] To recommend cloud computers, we need to calculate the relationship between the cloud computer to be recommended (i) and the user's interest vector. The third similarity between them. A commonly used similarity measure is cosine similarity.

[0104] Preferably, the third similarity Represented as:

[0105] (4);

[0106] in, The characteristics of cloud computers (i) are as follows: Let u be the user interest vector.

[0107] In the above scheme, by calculating the content similarity score based on the matching degree between the cloud computer's own features and the user's historical preferences, a stable recommendation dimension can be provided for the entire recommendation system, enabling the recommendation algorithm to understand the internal logic of user decision-making, rather than simply replicating historical behavior.

[0108] As a preferred implementation, step S4, obtaining the clustering feedback score for each user on each cloud computer based on the feedback scores of each user group on the configuration of each cloud computer, includes:

[0109] Obtain the first feedback score of each user in each user group on each of the cloud computers on the preset configuration items;

[0110] Based on the first feedback score, calculate the average feedback score for each user on all configuration items;

[0111] Based on the preset user weights and the average feedback score, the clustering feedback score of each user group for the configuration of each cloud computer is calculated.

[0112] Users within the same user group have similar characteristics. In this embodiment of the invention, the clustering feedback score of the user group for each cloud computer configuration is further calculated, which serves as a representative indicator of the overall clustering preference.

[0113] The first feedback score refers to the evaluation or satisfaction rating given by each individual user within the same user group for a specific configuration item of a cloud computer. This score reflects the user's subjective degree of approval for that configuration item.

[0114] The average feedback score is a value obtained by averaging the first feedback scores of all users in a user group for a cloud computer on each configuration item. It represents the overall evaluation level of the cloud computer by the group as a whole.

[0115] By considering weighted user feedback using preset user weights, the differences in importance among different user feedback can be reflected. In some preferred embodiments, user weights are determined based on user activity, trust level, and other relevant indicators. In other preferred embodiments, all users have a weight of 1 or a consistent non-negative value, indicating that there is no difference in importance among users.

[0116] For example, user The set of feedback ratings for cloud PC configurations is as follows: Then, the user group to which user u belongs The clustering feedback score is represented as:

[0117] (5);

[0118] in, The number of preset configuration items; The score for user u's first feedback on the i-th configuration item; User weight; This represents the average feedback score from user u for all configuration items.

[0119] The above scheme introduces two key mechanisms: preset configuration items and user weights. This ensures that the calculated clustering feedback score is both dimensionally refined and individually differentiated. Simultaneously, it leverages the collective preferences of other users within the group to provide initial recommendation criteria for that user, effectively mitigating the recommendation difficulties caused by the sparsity of individual data.

[0120] As a preferred implementation, step S5, calculating the recommendation score for each user for each of the cloud computers based on the collaborative filtering score, the content similarity score, and the clustering feedback score, includes:

[0121] Based on preset algorithm weights, the collaborative filtering score, the content similarity score, and the clustering feedback score are weighted and fused to obtain the recommendation score for each user for each cloud computer.

[0122] Based on the user's recommendation scores for all cloud computers, the N cloud computers with the highest recommendation scores are selected to generate cloud computer recommendation results for the user.

[0123] The preset algorithm weights refer to the pre-set coefficients used to control the degree of influence of the three scores in the final recommendation results. These weights can be set differently according to business objectives, user types or scenario requirements, and their specific values ​​do not affect the beneficial effects achieved by the embodiments of the present invention.

[0124] In some preferred embodiments, the recommendation score of user u for cloud computer i is represented as:

[0125] (6);

[0126] in, The collaborative filtering score for user u against cloud computer i; The content similarity score between user u and cloud computer i; The clustering feedback score for user u in user group k relative to cloud computer i; , and The preset weight parameters, .

[0127] It should be noted that the final cloud computer recommendation results presented to the user are an ordered list of cloud computers, and the order of the cloud computers in the recommendation results is determined by their recommendation scores.

[0128] In the above scheme, the cloud computer recommendation result integrates three aspects: collaborative filtering score, content similarity score, and clustering feedback score, forming a three-dimensional recommendation system that complements and verifies each other.

[0129] The cloud computer recommendation method provided by this invention, by simultaneously considering user behavior patterns, personal preferences and group consensus, can verify the rationality of recommendations from multiple perspectives, avoid the bias that may be caused by a single algorithm, and ensure that the recommendation results not only meet the user's personalized needs but are also verified by the group, thereby improving the user usability of the recommendation results.

[0130] This invention provides a cloud-based computer recommendation system. Please refer to [link / reference]. Figure 2 The cloud computer recommendation system includes a collaborative filtering algorithm application module 11, a user clustering module 12, a content recommendation algorithm application module 13, a clustering feedback application module 14, and an algorithm fusion module 15, wherein:

[0131] The collaborative filtering algorithm application module 11 is used to calculate the collaborative filtering score of each user for each cloud computer based on the user's historical behavior records of the cloud computer using a collaborative filtering algorithm.

[0132] User clustering module 12 is used to extract user features and cluster users according to the user features to obtain several user groups;

[0133] The content recommendation algorithm application module 13 is used to extract cloud computer features and, based on the cloud computer features, use a content-based recommendation algorithm to calculate the content similarity score of each user to each cloud computer.

[0134] Clustering feedback application module 14 is used to obtain the clustering feedback score of each user for each cloud computer based on the feedback score of each user group for the configuration of each cloud computer.

[0135] The algorithm fusion module 15 is used to calculate the recommendation score for each user for each of the cloud computers based on the collaborative filtering score, the content similarity score, and the clustering feedback score.

[0136] In a preferred embodiment, the collaborative filtering algorithm application module 11 is specifically used for:

[0137] Based on users' historical behavior records of cloud computers, calculate the first similarity between users and the second similarity between cloud computers;

[0138] Based on the first similarity, the second similarity, and the user's rating behavior for historical interactive cloud computers, the predicted rating of the user for all cloud computers is calculated, and the collaborative filtering score of each user for each cloud computer is obtained.

[0139] In a preferred embodiment, the user clustering module 12 is specifically used for:

[0140] Extract user characteristics;

[0141] Based on the user characteristics, the ant colony algorithm is used to obtain the initial cluster centers of the users; wherein, the mutation factor of the ant colony algorithm is defined using the golden sine algorithm.

[0142] Based on the initial cluster centers and the user characteristics, the K-means clustering algorithm is used to cluster the users, resulting in several user groups.

[0143] In a preferred embodiment, the content recommendation algorithm application module 13 is specifically used for:

[0144] Extracting features from cloud computers;

[0145] Based on the cloud computer characteristics described in the user's historical interactions with the cloud computer, construct a user interest vector;

[0146] Calculate the third similarity between the cloud computer features of each cloud computer and the user interest vector to obtain the content similarity score of each user for each cloud computer.

[0147] In a preferred embodiment, the clustering feedback application module 14 is specifically used for:

[0148] Obtain the first feedback score of each user in each user group on each of the cloud computers on the preset configuration items;

[0149] Based on the first feedback score, calculate the average feedback score for each user on all configuration items;

[0150] Based on the preset user weights and the average feedback score, the clustering feedback score of each user group for the configuration of each cloud computer is calculated.

[0151] In a preferred embodiment, the algorithm fusion module 15 is specifically used for:

[0152] Based on preset algorithm weights, the collaborative filtering score, the content similarity score, and the clustering feedback score are weighted and fused to obtain the recommendation score for each user for each cloud computer.

[0153] Based on the user's recommendation scores for all cloud computers, the N cloud computers with the highest recommendation scores are selected to generate cloud computer recommendation results for the user.

[0154] The cloud computer recommendation system provided by this invention, by simultaneously considering user behavior patterns, personal preferences, and group consensus, can verify the rationality of recommendations from multiple perspectives, avoiding the bias that may be caused by a single algorithm, and ensuring that the recommendation results not only meet the user's personalized needs but are also verified by the group, thereby improving the user usability of the recommendation results.

[0155] Please see Figure 3 , Figure 3 This is a structural block diagram of a cloud computer recommendation device provided in an embodiment of the present invention. The cloud computer recommendation device includes a processor 31, a memory 32, and a computer program stored in the memory 32 and executable on the processor 31. When the processor 31 executes the computer program, it implements the steps in the various cloud computer recommendation method embodiments described above, such as steps S1 to S5.

[0156] For example, the computer program can be divided into one or more modules / units, which are stored in the memory 32 and executed by the processor 31 to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the cloud computer recommendation device.

[0157] The cloud computer recommended device may include, but is not limited to, processor 31 and memory 32. Those skilled in the art will understand that the schematic diagram is merely an example of a cloud computer recommended device and does not constitute a limitation on the cloud computer recommended device. It may include more or fewer components than illustrated, or combine certain components, or different components. For example, the cloud computer recommended device may also include input / output devices, network access devices, buses, etc.

[0158] The processor 31 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor 31 is the control center of the cloud computer recommendation device, connecting various parts of the entire cloud computer recommendation device through various interfaces and lines.

[0159] The memory 32 can be used to store the computer programs and / or modules. The processor 31 implements various functions of the cloud computer recommended device by running or executing the computer programs and / or modules stored in the memory 32 and calling the data stored in the memory 32. The memory 32 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 32 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0160] Wherein, if the modules / units integrated into the cloud computer recommended device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor 31, it can implement the steps of the various method embodiments described above. Wherein, the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0161] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A cloud computer recommendation method, characterized in that, include: Based on the user's historical behavior records of the cloud computers, a collaborative filtering algorithm is used to calculate the collaborative filtering score for each user on each of the cloud computers. Extract user features, and cluster users based on these features to obtain several user groups; Extract cloud computer features, and based on the cloud computer features, use a content-based recommendation algorithm to calculate the content similarity score for each user to each cloud computer; Based on the feedback scores of each user group on the configuration of each cloud computer, the clustering feedback score of each user on each cloud computer is obtained. Based on the collaborative filtering score, the content similarity score, and the clustering feedback score, calculate the recommendation score for each user for each of the cloud computers; Based on users' historical behavior records with cloud computers, a collaborative filtering algorithm is used to calculate the collaborative filtering score for each user on each of the aforementioned cloud computers, including: Based on users' historical behavior records of cloud computers, calculate the first similarity between users and the second similarity between cloud computers; Based on the first similarity, the second similarity, and the user's rating behavior for historical interactive cloud computers, the predicted rating of the user for all cloud computers is calculated, and the collaborative filtering score of each user for each cloud computer is obtained. The process involves extracting cloud computer features, and based on these features, employing a content-based recommendation algorithm to calculate the content similarity score for each user across all cloud computers, including: Extracting features from cloud computers; Based on the cloud computer characteristics described in the user's historical interactions with the cloud computer, construct a user interest vector; Calculate the third similarity between the cloud computer features of each cloud computer and the user interest vector to obtain the content similarity score of each user for each cloud computer; The step of obtaining the clustering feedback score for each user on each cloud computer based on the feedback scores of each user group on the configuration of each cloud computer includes: Obtain the first feedback score of each user in each user group on each of the cloud computers on the preset configuration items; Based on the first feedback score, calculate the average feedback score for each user on all configuration items; Based on the preset user weights and the average feedback score, the clustering feedback score of each user group for the configuration of each cloud computer is calculated.

2. The cloud computer recommendation method as described in claim 1, characterized in that, The step involves extracting user features and clustering users based on these features to obtain several user groups, including: Extract user characteristics; Based on the user characteristics, the ant colony algorithm is used to obtain the initial cluster centers of the users; wherein, the mutation factor of the ant colony algorithm is defined using the golden sine algorithm. Based on the initial cluster centers and the user characteristics, the K-means clustering algorithm is used to cluster the users, resulting in several user groups.

3. The cloud computer recommendation method as described in claim 1, characterized in that, The step of calculating the recommendation score for each user for each of the cloud computers based on the collaborative filtering score, the content similarity score, and the clustering feedback score includes: Based on preset algorithm weights, the collaborative filtering score, the content similarity score, and the clustering feedback score are weighted and fused to obtain the recommendation score for each user for each cloud computer. Based on the user's recommendation scores for all cloud computers, the N cloud computers with the highest recommendation scores are selected to generate cloud computer recommendation results for the user.

4. A cloud-based computer recommendation system, characterized in that, include: The collaborative filtering algorithm application module is used to calculate the collaborative filtering score for each user on each cloud computer based on the user's historical behavior records on the cloud computers using a collaborative filtering algorithm. The user clustering module is used to extract user features and cluster users according to the user features to obtain several user groups; The content recommendation algorithm application module is used to extract cloud computer features and, based on the cloud computer features, employ a content-based recommendation algorithm to calculate the content similarity score for each user to each of the cloud computers. The clustering feedback application module is used to obtain the clustering feedback score of each user for each cloud computer based on the feedback score of each user group for the configuration of each cloud computer. The algorithm fusion module is used to calculate the recommendation score for each user for each of the cloud computers based on the collaborative filtering score, the content similarity score, and the clustering feedback score. The collaborative filtering algorithm application module is specifically used for: Based on users' historical behavior records of cloud computers, calculate the first similarity between users and the second similarity between cloud computers; Based on the first similarity, the second similarity, and the user's rating behavior for historical interactive cloud computers, the predicted rating of the user for all cloud computers is calculated, and the collaborative filtering score of each user for each cloud computer is obtained. The user clustering module is specifically used for: Extract user characteristics; Based on the user characteristics, the ant colony algorithm is used to obtain the initial cluster centers of the users; wherein, the mutation factor of the ant colony algorithm is defined using the golden sine algorithm. Based on the initial cluster centers and the user characteristics, the K-means clustering algorithm is used to cluster the users, resulting in several user groups; The clustering feedback application module is specifically used for: Obtain the first feedback score of each user in each user group on each of the cloud computers on the preset configuration items; Based on the first feedback score, calculate the average feedback score for each user on all configuration items; Based on the preset user weights and the average feedback score, the clustering feedback score of each user group for the configuration of each cloud computer is calculated.

5. A cloud computer recommendation device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the cloud computer recommendation method as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the cloud computer recommendation method as described in any one of claims 1 to 3.

7. A computer program product, characterized in that, The computer program product includes a computer program or computer instructions, which, when executed by a processor, perform the cloud computer recommendation method as described in any one of claims 1 to 3.

Images