A game user loss early warning method and system

Abstract

The application relates to a game user loss early warning method and system, and relates to the field of game user information processing technology, which comprises the following steps: step 1, obtaining friend game data of a game user; step 10, extracting the friend game data based on double-party interaction information to form a social circle set; step 11, screening the number and type of friend loss in the social circle set; step 12, searching for a corresponding calculation ratio based on the type of friend loss; step 13, calculating a friend loss value according to a preset weighting sum formula based on the number of friend loss of the type of friend loss and the corresponding calculation ratio of the type of friend loss; and step 14, when the friend loss value is greater than a preset friend loss early warning value, marking the game user as a high-risk loss user and outputting a user loss early warning signal. The application achieves the effect of predicting the loss of the game user based on the same social circle, and improves the accuracy of the game user loss early warning.

Description

Technical Field

[0001] This invention relates to the field of game user information processing technology, and in particular to a method and system for early warning of game user churn. Background Technology

[0002] Game user churn warning is the process of identifying users with a tendency to churn by analyzing relevant data and sending early warning signals before users officially leave the game. User churn warning has significant strategic importance for the operation and development of game platforms.

[0003] In related technologies, the game user churn warning method mainly involves statistically analyzing the user behavior of each game user on the game platform based on the game duration, comparing the statistical results of user behavior with preset behavior thresholds, and generating retention or recall strategies when the statistical results of user behavior exceed the preset behavior thresholds, and pushing them to the corresponding user terminals.

[0004] Regarding the aforementioned technologies, the inventors believe that existing prediction methods are tailored to a single user, relying solely on their login time. The data is relatively limited, and existing methods and systems have limited ability to predict the churn tendency of game users, leaving room for improvement. Summary of the Invention

[0005] To improve the accuracy of predicting game user churn, this invention provides a game user churn early warning method and system that flexibly implements corresponding recall strategies based on the same user's social and gaming groups and provides effective feedback to the recall strategies.

[0006] In a first aspect, the present invention provides a method for early warning of game user churn, which adopts the following technical solution:

[0007] A method for early warning of game user churn includes:

[0008] Step 1: Obtain the game data of the game user's friends, which includes the interaction information between the game user and their friends and the friend's personal login information;

[0009] Step 10: Extract game data from friends based on mutual interaction information to form a social circle set;

[0010] Step 11: Filter the number and type of friends lost within the social circle;

[0011] Step 12: Find the corresponding calculation ratio based on the type of friend churn;

[0012] Step 13: Calculate the friend churn value based on the number of friends churned according to the type of friend churn and the calculation ratio corresponding to the type of friend churn according to the preset weighted sum formula;

[0013] Step 14: When the friend churn value exceeds the preset friend churn warning value, mark the game user as a high-risk churn user and output a user churn warning signal;

[0014] Among them, the methods for predicting friend churn after task release based on friend data include:

[0015] Step 110: Obtain the expected release information and expected release time for the task;

[0016] Step 111: Based on the social circle set, obtain the historical shared task types, historical shared task data, and historical login information of the user and friends;

[0017] Step 112: Input the historical common task types, historical common task data and historical login information into the preset standard simulation model for correction to obtain the adapted simulation model, and analyze the expected task release information to obtain the expected task type;

[0018] Step 113: Input the expected task type into the adaptation simulation model to obtain the expected common task data and expected login information;

[0019] Step 114: Calculate the attraction value based on the expected release time, expected shared task data, and expected login information, and update the friend churn value according to the attraction value to obtain the expected friend churn value.

[0020] By adopting the above technical solution, the system was able to filter out friends who have close interactions with users in the game, achieving the effect of predicting game user churn based on the same social circle, and improving the accuracy of game user churn warning.

[0021] Optionally, specific methods for extracting game data from friends based on mutual interaction information to form social circles include:

[0022] Step 2: Obtain shared play information between game users and their friends based on social circles. The shared play information includes shared play data types and shared play data values.

[0023] Step 20: Find the corresponding intimacy threshold based on the data types of games played together;

[0024] Step 21: If there is at least one shared play data type with a shared play data value greater than the intimacy threshold, then extract the friend's game data into the social circle set;

[0025] Step 22: If the shared play data values ​​corresponding to all shared play data types are less than the intimacy threshold, then the friend's game data will not be included in the social circle set.

[0026] By adopting the above technical solution, the system can filter out friends who interact closely with users within the same social group based on the interaction between game users and their friends in the game, thus solving the problem of predicting the impact of close friends within the same social group on user churn.

[0027] Optionally, optimization methods for calculating friend churn values ​​based on the number of friends churned according to the type of friend churn and the calculation ratio corresponding to the type of friend churn, according to a preset weighted sum formula, include:

[0028] Step 3: Define the shared play data types in the friend game data of the social circle that meet the condition that the shared play data value is greater than the intimacy threshold as feature data types;

[0029] Step 30: Calculate the difference between the shared play data value and the intimacy threshold corresponding to the feature data type;

[0030] Step 31: Convert the difference into the increase in the calculated proportion of the friend churn data according to the preset formula for calculating the excess portion;

[0031] Step 32: Adjust the calculation ratio according to the increase to obtain a higher calculation ratio;

[0032] Step 33: After updating the high calculation ratio to the calculation ratio corresponding to the friend churn type, calculate the friend churn value according to the preset weighted sum calculation formula.

[0033] By adopting the above technical solution, the calculation weight of friend game data corresponding to friends who have close interactions with game users is increased, thereby calculating the friend churn value, improving the reliability of friend churn value calculation, and improving the accuracy of game user churn prediction based on social circles.

[0034] Optional, also includes:

[0035] Step 4: Based on the user's game data, find the user's and their friends' game chat data, and obtain specific vocabulary information based on the game chat data;

[0036] Step 40: Define the game users where the specific word information is located as circle-based early warning users, and the circle-based early warning users are users who have not yet churned;

[0037] Step 41: Calculate the number and type of churn for users who received alerts, and update the data to the number and type of friend churn.

[0038] Step 42: If there are preset high-risk words in the specific word information, then define the users in the circle where the high-risk words are located as high-risk users in the circle.

[0039] Step 43: Count the number and type of high-risk churn corresponding to high-risk users in the circle. Based on the number of high-risk churns and the number of churns under warning, calculate the high-risk ratio value according to the preset high-risk ratio calculation formula.

[0040] Step 44: Convert the high-risk ratio value according to the preset high-risk ratio calculation formula to obtain the increase ratio of the calculated ratio of the number of high-risk losses;

[0041] Step 45: Adjust the calculation ratio according to the increase ratio to obtain the high-risk calculation ratio and update it to the calculation ratio.

[0042] By adopting the above technical solution, the game chat data of users and their friends in the social circle can be found through friend game data, so as to realize the churn prediction of friends who have specific words in the chat, and further improve the accuracy of the user churn warning when specific words appear in the chat.

[0043] Optional, also includes:

[0044] Step 5: Obtain shared task data from non-social circles and non-friends;

[0045] Step 50: Determine the judgment parameters based on non-friend shared task data;

[0046] Step 51: Filter the common task data type in the friend churn type and the friend task data corresponding to the common task data;

[0047] Step 52: Determine the corresponding friend judgment threshold based on the shared task data of friends;

[0048] Step 53: If the judgment parameter is greater than the friend judgment threshold, the non-friend shared task data is regarded as friend game data and extracted into the social circle set.

[0049] By adopting the above technical solution, and by acquiring, filtering, and judging the data of non-friends on shared tasks, the problem of non-friends also affecting the online status of game users but not entering the social circle for churn prediction is solved.

[0050] Optionally, if the non-friend shared task data is treated as friend game data and extracted into the social circle, the specific method for calculating the friend churn value based on the number of friends churned according to the friend churn type and the calculation ratio corresponding to the friend churn type according to the preset weighted sum formula includes:

[0051] Step 6: Classify the non-friend shared task data into high-efficiency task churn type, low-efficiency task churn type, and composite task churn type according to the corresponding shared task threshold;

[0052] Step 60: Calculate the excess value of the specific task data corresponding to the high-efficiency task churn type, low-efficiency task churn type, and compound task churn type and the corresponding common task threshold using the difference calculation method;

[0053] Step 61: Based on the excess value, calculate the specific increase in the calculation ratio for the specific task according to the preset task churn excess value calculation formula;

[0054] Step 62: Update the calculation ratio based on the specific increase to obtain the specific task calculation ratio.

[0055] By adopting the above technical solution, the common task data of non-friend users and game users can be classified, and the impact of non-friend users on game users' gameplay can be classified and calculated, thereby improving the accuracy of user churn warning.

[0056] Optional, also includes:

[0057] Step 7: In response to the friend recall task sent to friends who have been lost in the social circle, obtain the recall task type and corresponding friend task data after the preset warning period. The friend task data includes the number of friend tasks, the number of friend tasks completed, the friend task time, and the friend return time, etc.

[0058] Step 70: Analyze the upward trend of friend return time based on the preset early warning period;

[0059] Step 71: When the upward trend is 0, no action is taken;

[0060] Step 72: When the upward trend is greater than 0 and the friend return time is still less than the preset full return threshold, calculate and analyze the friend churn task completion rate based on the number of friend task completions and the number of friend task completions.

[0061] Step 73: When the completion rate of the friend churn task is greater than the preset friend churn task completion threshold, remove the friend from the number of friends churned in the social circle.

[0062] Step 74: When the completion rate of the friend churn task is less than the friend churn task completion threshold, replace the friend recall task from the preset task list.

[0063] Step 75: When the upward trend is greater than 0 and the friend's return time is greater than the full return threshold, remove the friend from the number of friends lost in the social circle.

[0064] By adopting the above technical solution, the return of friends can be analyzed based on the task data of friends, making the game data of friends in the social circle more complete and forming an effective feedback for the calculation of friend churn value.

[0065] Optionally, it also includes a method to change the friend recall task from the preset task list:

[0066] Step 8: Define friend recall tasks with a friend churn task completion rate lower than the friend churn task completion threshold as failed recall tasks;

[0067] Step 80: Determine the corresponding task data threshold based on the recall task type. The task data threshold includes the friend task time threshold and the friend task completion threshold.

[0068] Step 81: When the friend task time exceeds the friend task time threshold and the number of friend tasks completed exceeds the friend task completion threshold, a high-difficulty task is matched in the recall task library based on the recall task type, and the friend recall task is replaced.

[0069] Step 82: When the friend task time exceeds the friend task time threshold and the number of friend tasks completed is less than the friend task completion threshold, a low-difficulty task is matched in the recall task library based on the recall task type, and the friend recall task is replaced.

[0070] By adopting the above technical solution, the problem of the inability of the friend churn value to provide effective feedback to the subsequent recall strategy is solved by determining the type of recall task and changing the recall task based on the completion rate of the friend churn task.

[0071] Secondly, this invention provides a game user churn early warning system, which adopts the following technical solution:

[0072] A game user churn early warning system includes:

[0073] The acquisition module is used to acquire game data from friends, game chat data, shared task data from non-friends, expected task release information, expected release time, historical shared task types, historical shared task data, historical login information, and friend task data.

[0074] A memory for storing a program for a game user churn early warning method as described above;

[0075] The processor loads and executes programs from memory.

[0076] By adopting the above technical solution, the acquisition of various data and the calculation of friend churn values ​​were realized, and the prediction of game user churn was achieved based on the churn of friends within the same social group.

[0077] In summary, the present invention has at least one of the following beneficial technical effects:

[0078] 1. It enables the filtering of friends who have close interactions with users in the game, achieving the effect of predicting game user churn based on the same social circle, and improving the accuracy of game user churn warning;

[0079] 2. It enables the prediction of user return and churn rates when tasks are released, expanding the scope of user churn warnings;

[0080] 3. It has achieved the acquisition of various data and the calculation of friend churn values, and has achieved the effect of predicting the churn of game users based on the churn of friends within the same social group. Attached Figure Description

[0081] Figure 1 This is a flowchart of a method for predicting user churn in games.

[0082] Figure 2 This is a flowchart illustrating a specific method for extracting game data from friends based on information exchanged between the two parties to form a social circle.

[0083] Figure 3 This is a flowchart of an optimization method based on the number of friends lost and the calculation ratio corresponding to the type of friend churn.

[0084] Figure 4 This is a flowchart for finding game chat data of users and their friends within their social circles based on their game data.

[0085] Figure 5 This is a schematic diagram of a game user churn early warning system. Detailed Implementation

[0086] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0087] This invention discloses a method for early warning of user churn in games. (Refer to...) Figure 1 One method for predicting game user churn includes:

[0088] Step 1: Obtain the game data of the game user's friends.

[0089] Friend game data refers to various data about the game friends added by the game user. This includes interaction information between the game user and the friend, as well as the friend's personal login information. Specifically, it includes information on the intimacy level between the game user and the friend, friend time information, friend chat information, shared play information, shared task data, and login data. Friend game data is obtained by searching various data of game friends in the game's backend database.

[0090] Step 10: Extract game data from friends based on the interaction information between the two parties to form a social circle.

[0091] The social circle dataset is a dataset that filters all friend game data, retaining only the game data of friends with high levels of interaction with the user. The extraction method will be introduced later and will not be elaborated here.

[0092] Step 11: Filter the number and type of friends lost in the social circle.

[0093] Friend churn type refers to the type of reason for friend churn within a social circle, including churn due to win rate, IP address, shared playtime, shared quests, and shared items. Friend churn quantity refers to the number of friends churned for each churn type.

[0094] Filter by friend churn type and count the number of friends churned for each type.

[0095] Step 12: Find the corresponding calculation ratio based on the type of friend churn.

[0096] The calculation ratio is the weighted sum of the number of friends churned for each type of friend churn. When searching for the calculation ratio, it should satisfy... W i Calculate the proportion of various types of friend churn.

[0097] Step 13: Calculate the friend churn value based on the number of friends churned according to the type of friend churn and the calculation ratio corresponding to the type of friend churn, according to the preset weighted sum formula.

[0098] The friend churn value is the probability value that the loss of a game user's friends will affect the user's churn.

[0099] The formula for calculating the weighted sum of friend churn value is as follows: Where V is the friend churn value, S i The number of friends lost.

[0100] Step 14: When the friend churn value exceeds the preset friend churn warning value, mark the game user as a high-risk churn user and output a user churn warning signal.

[0101] The friend churn warning value is a threshold value set by professionals in this field. It indicates that the loss of close friends poses a risk of user churn. When this threshold is exceeded, a warning signal is issued. For example, assuming the data for various friend churn types are S1=0.4, S2=0.2, S3=0.9, S4=0.6, and the calculation ratios for each type are W1=0.15, W2=0.1, W3=0.45, W4=0.3, the pre-defined weighted sum formula yields V=0.665. The pre-defined friend churn warning value is 0.55. If the friend churn value exceeds this threshold, it indicates that the loss of close friends poses a risk of user churn, and the user is marked as a high-risk churner.

[0102] High-risk churn users are those whose churn is attributed to the loss of their friends, thus raising the risk of further churn. The user churn warning signal is an alert that a user may churn when the number of friends lost exceeds a preset warning threshold.

[0103] Step 110: Obtain the expected release information and expected release time for the task.

[0104] The expected release information for a task includes details such as the task type, rewards, difficulty, and opening time window. The expected release time is the specific time the task is expected to be released. This information is retrieved from pre-stored, unreleased task content in the game's backend.

[0105] Step 111: Based on the social circle, obtain the historical shared task types, historical shared task data, and historical login information of the user and friends.

[0106] The "Historical Shared Tasks" type refers to past shared task data. "Historical Shared Tasks Data" refers to past shared task data corresponding to the previous shared task data type. "Historical Login Information" includes information such as the number of days a user and their friends have logged in synchronously and the number of days they have been online together. The method for obtaining this information is similar to step 2.

[0107] Step 112: Input the historical common task types, historical common task data and historical login information into the preset standard simulation model for correction to obtain the adapted simulation model, and analyze the expected task release information to obtain the expected task type.

[0108] The standard simulation model is a general model that simulates the level of player interaction during tasks, and it is pre-set based on past player task interaction data. The adaptive simulation model is a simulation model that is modified by inputting historical common task types, historical common task data, and historical login information into the standard simulation model so that it can predict the type of tasks to be released.

[0109] The expected task type refers to the type of task that is expected to be released. The expected task type is obtained by classifying and analyzing the historical tasks that have been released, such as multiplayer team type, single-player task type, guild battle type, etc.

[0110] Step 113: Input the expected task type into the adaptation simulation model to obtain the expected common task data and expected login information.

[0111] The expected shared task data refers to the shared task data that game users and their friends are expected to see after the task is released. The expected login information refers to the login information of users and their friends expected after the task is completed. By inputting the expected task type into the adaptation simulation model, the adaptation simulation model simulates and calculates the expected task type based on the already input historical shared task types, historical shared task data, and historical login information to obtain the expected shared task data and expected login information.

[0112] Step 114: Calculate the attraction value based on the expected release time, expected shared task data, and expected login information, and update the friend churn value according to the attraction value to obtain the expected friend churn value.

[0113] Attraction value is a calculated numerical value that estimates the return or play rate of friends and users after the task is released. The calculated attraction value meets the following criteria: Where Va is the attraction value, L is the expected release time, K is the expected shared task data, and D is the expected login data.

[0114] The expected friend churn value is a churn value that is updated based on the attraction value to predict the churn rate of friends and users after the task is released.

[0115] Reference Figure 2 Specific methods for extracting game data from friends based on mutual interaction information to form social circles include:

[0116] Step 2: Obtain information on shared gameplay among game users and their friends based on social circles.

[0117] Shared play information refers to various data and corresponding types that users and their friends use when playing games together, including shared play data types and shared play data values. Shared play information is extracted from friends' game data within a social circle.

[0118] Step 20: Find the corresponding intimacy threshold based on the data type of shared play.

[0119] The intimacy threshold is a threshold that determines the relationship between a friend and a game user when a shared gameplay data value exceeds the threshold for judging the friend's online gaming status and its significant impact on the game user. Shared gameplay data types refer to various aspects of gameplay between the game user and friend, including shared game wins, game IP addresses, shared game time, shared game counts, and shared item acquisitions. Shared gameplay data values ​​are the corresponding data values ​​for each type of gameplay, including shared game win counts, game IP address data, shared game time values, shared game counts, and shared item acquisition values. The intimacy threshold is determined based on the shared gameplay data types, and the value of the intimacy threshold is set by past shared gameplay data values, such as determining the shared game time threshold based on the shared game time data type.

[0120] Step 21: If there is at least one shared play data type with a shared play data value greater than the intimacy threshold, then extract the friend's game data into the social circle set.

[0121] When the value of at least one shared play data type exceeds the intimacy threshold, it indicates that the shared play data value between the friend and the game user exceeds the value used to determine the relationship between the two, and the friend's online game status has a significant impact on the game user. In this case, the friend's game data is extracted into the social circle set for calculating friend churn value. For example, when the number of times the user and a game friend play together exceeds the shared play threshold, the friend's game data is extracted into the social circle set.

[0122] Step 22: If the value of the shared play data corresponding to all shared play data types is less than the intimacy threshold, then the friend's game data will not be included in the social circle set.

[0123] If the shared play data values ​​corresponding to all shared play data types are less than the intimacy threshold, it means that no single shared play data value between the friend and the game user exceeds the value used to determine the relationship between the two. The friend's online game status has no impact on the game user, so the friend's game data is not included in the social circle set.

[0124] Reference Figure 3 The optimization method for calculating the friend churn value based on the number of friends churned according to the type of friend churn and the calculation ratio corresponding to the type of friend churn, according to a preset weighted sum formula, includes:

[0125] Step 3: Define the shared play data types in the friend game data of the social circle that meet the condition that the shared play data value is greater than the intimacy threshold as feature data types.

[0126] Step 30: Calculate the difference between the shared play data value and the intimacy threshold corresponding to the feature data type.

[0127] The difference is the difference between the shared play data value corresponding to the feature data type and the intimacy threshold.

[0128] When calculating the difference between the shared play data value and the intimacy threshold, the following conditions must be met: and ,in The difference between the shared play data values ​​and the intimacy threshold. For each feature data type, there are common play data values. The proximity threshold is the corresponding threshold for each feature data type.

[0129] Step 31: Convert the difference into the increase in the calculated proportion of the friend churn data according to the preset formula for calculating the excess portion.

[0130] The increase is the percentage increase calculated using a weighted sum formula based on the difference between shared play data values ​​and the intimacy threshold. When calculating the increase, the preset formula for calculating the excess portion is as follows: ,in The percentage increase in the calculated proportion of various friend churn types. The conversion coefficients corresponding to various types of friend churn are preset, and the impact of friend churn type on the early warning is pre-set in the program.

[0131] Step 32: Adjust the calculation ratio according to the increase to obtain a high calculation ratio.

[0132] The high calculation ratio is the calculation ratio corresponding to various friend churn types after adjustment based on the increase magnitude. When adjusting the calculation ratio, the formula is used... The corresponding calculation ratios are corrected, among which... For high computational ratios, this process is completed before step 13.

[0133] Step 33: After updating the high calculation ratio to the calculation ratio corresponding to the friend churn type, calculate the friend churn value according to the weighted sum calculation formula.

[0134] Reference Figure 4 It also includes:

[0135] Step 4: Based on the game data of friends, find the game chat data of the user and friends in the social circle, and find the preset specific word information based on the game chat data.

[0136] Game chat data consists of records of interactions between users and their friends within the game's user data. Specific vocabulary information refers to words that indicate a user's or friend's churn intention. This information is pre-configured by professionals based on past game churn patterns; for example, words like "quit," "leave the game," and "sell account" are identified as specific vocabulary information. Game chat data is obtained by searching the backend database of user and friend chat data. Finding specific vocabulary information involves comparing the collected game chat data with a pre-configured specific vocabulary database to identify data containing specific vocabulary information.

[0137] Step 40: Define the game users whose specific vocabulary information is located as users of the circle alert.

[0138] The alert targets users who have appeared in social circles using specific keywords but have not yet been churned.

[0139] Step 41: Calculate the number and type of churn for each user who received a warning, and update the data to the number and type of friend churn.

[0140] The "Warning Churn Count" refers to the number of friends lost by the user who received the warning. The "Warning Churn Type" refers to the type of friend loss experienced by the user who received the warning.

[0141] Step 42: If there are preset high-risk words in the specific word information, then define the users in the circle where the high-risk words are located as high-risk users in the circle.

[0142] High-risk keyword information refers to sensitive keywords within specific keyword information that indicate a strong tendency to churn friends and users. The pre-setting method is similar to that for specific keyword information; for example, the presence of words like "permanent uninstall" and "account deletion" is considered specific keyword information. High-risk user targeting refers to users identified through high-risk keyword information in game chat data who are flagged for potential detection.

[0143] Step 43: Count the number and type of high-risk churn corresponding to high-risk users, and calculate the high-risk ratio based on the number of high-risk churns and the number of churns under warning.

[0144] The "high-risk churn count" refers to the number of friends lost by users identified as high-risk. The "high-risk churn type" refers to the type of friend churn experienced by users identified as high-risk. The formula for calculating the high-risk ratio is as follows: ,in The proportion of high-risk individuals is [value missing]. For high-risk losses, To provide early warning of the number of losses. The larger the value, the higher the proportion of the population with a tendency to churn, and the dimensional weight needs to be increased.

[0145] Step 44: Convert the high-risk ratio value according to the preset high-risk ratio calculation formula to obtain the increase ratio of the calculated high-risk loss quantity.

[0146] The high-risk ratio calculation formula is a formula that converts the high-risk ratio value into an increased high-risk calculation ratio. The high-risk calculation ratio is the calculated ratio corresponding to various early warning churn types after adjusting for the number of high-risk churns. The high-risk ratio calculation formula is as follows: ,in The percentage increase is calculated based on the number of high-risk churns. A is the warning amplification coefficient, which is preset by domain personnel based on past warning user churn data to limit the extent of the percentage increase.

[0147] Step 45: Adjust the calculation ratio according to the increase ratio to obtain the high-risk calculation ratio and update it to the calculation ratio.

[0148] The revised high-risk calculation ratio should meet the following requirements. ,in The proportion is calculated for high-risk situations. For example, when... , At that time, set ,at this time The high-risk calculation ratio The value is 0.585. This process is completed before step 13.

[0149] This also includes:

[0150] Step 5: Obtain shared task data from non-social circles and non-friends.

[0151] The non-social circle set is a dataset of game data excluded from the social circle set. Non-friend collaborative task data consists of data on tasks completed jointly by game users and non-friend players, including the number of collaborative tasks, duration of collaborative tasks, interactive items used in tasks, and the number of collaborative tasks completed. This non-social circle data for non-friend collaborative tasks is obtained by searching the game's backend database for various data related to interactions between in-game friends and non-friends.

[0152] Step 50: Determine the judgment parameters based on non-friend shared task data.

[0153] The judgment parameters are various parameters used to measure the level of task participation by unfamiliar players and users, determined based on shared task data among non-friends. The determination of these parameters was accomplished by analyzing shared task data among non-friends and referencing interaction data between users and friends, and between users and non-friends, during game tasks.

[0154] Step 51: Filter the shared task data type and the corresponding friend task data in the friend churn type.

[0155] Friend task data refers to the shared tasks performed by friends within a social circle. Filtering friend task data determines whether the effectiveness of shared tasks performed by non-friends is the same as that performed by friends within the social circle, thus confirming that non-friends also contribute to user churn. This is achieved by comparing the number of churned users for shared tasks within the friend churn categories to filter the relevant friend task data.

[0156] Step 52: Determine the corresponding friend judgment threshold based on the shared task data of friends.

[0157] The friend selection threshold is a critical value derived from shared task data with friends, used to determine whether to include non-friend shared task data within the social circle. The value of the friend selection threshold is determined by past shared play data with friends, such as determining the threshold based on the shared task time.

[0158] Step 53: If the judgment parameter is greater than the friend judgment threshold, the non-friend shared task data is regarded as friend game data and extracted into the social circle set.

[0159] When the judgment parameter is greater than the friend judgment threshold, it indicates that the non-friend also has close interaction with the user, and the impact on user churn is the same as that of a friend. In this case, the shared task data of the non-friend is regarded as friend game data and extracted into the social circle set. This process is completed before step 13.

[0160] Specifically, if the non-friend shared task data is treated as friend game data and extracted into the social circle, the specific method for calculating the friend churn value based on the number of friends churned by the friend churn type and the calculation ratio corresponding to the friend churn type according to the preset weighted sum formula includes:

[0161] Step 6: Classify the non-friend shared task data into high-efficiency task churn type, low-efficiency task churn type, and composite task churn type according to the corresponding shared task threshold.

[0162] The shared task threshold is the threshold for different types of shared tasks between non-friends and users when they team up to perform tasks.

[0163] High-efficiency task churn type refers to a data type of collaborative task that is highly efficient when non-friends team up with other users. This type of data is classified as high-efficiency task churn when both the number of collaborative tasks performed and the number of completed collaborative tasks exceed a threshold.

[0164] Inefficient task churn type refers to non-friends' shared task data where the completion efficiency is low when teaming up with other users. When the number of shared tasks in the non-friends' shared task data is less than the task count threshold and the number of completed shared tasks is greater than the shared task completion threshold, the non-friends' shared task data is classified as inefficient task churn type.

[0165] Composite task churn type refers to a non-friend shared task data type where all shared task data exceeds the corresponding task threshold when non-friends and users team up to perform tasks.

[0166] Step 60: Calculate the excess value of the specific task data corresponding to the high-efficiency task churn type, low-efficiency task churn type, and compound task churn type and the corresponding common task threshold using the difference calculation method.

[0167] The difference calculation method is used to calculate the excess value of a common task threshold for task data of high-efficiency task churn, low-efficiency task churn, and combined task churn. When calculating the difference calculation method for high-efficiency and low-efficiency task churn, the calculation formula is as follows: ,in This represents the value exceeding the corresponding common task threshold. This provides specific task data corresponding to high-efficiency and low-efficiency task churn patterns. This is the threshold for common tasks. When calculating the churn rate for composite tasks using the difference calculation method, the formula is as follows: ,in This refers to the specific task data corresponding to the composite task loss type.

[0168] Step 61: Based on the excess value, calculate the specific increase in the calculation ratio for the specific task according to the preset task churn excess value calculation formula.

[0169] The specific increase is the increase that should be updated in the calculation ratio, calculated according to the task churn excess value calculation formula after classifying non-friend task data. The task churn excess value calculation formula is a formula used to calculate the specific increase in the calculation ratio for each specific task after classification of non-friend task data. When calculating the specific increase, the following conditions must be met: , The specific increase is indicated below. The enhancement coefficients for each pre-set task type are determined based on past data from non-friend tasks.

[0170] Step 62: Update the calculation ratio based on the specific increase to obtain the specific task calculation ratio.

[0171] The formula for updating the calculation ratio to obtain the calculation ratio for a specific task is as follows: W tThis refers to the specific task calculation ratio. This process is completed before step 13.

[0172] This also includes:

[0173] Step 7: In response to the friend recall task sent to friends who have been lost in the social circle, obtain the recall task type and corresponding friend task data after the preset warning period.

[0174] A friend recall task is a task issued to estranged friends within a social circle to encourage them to return. The task type refers to the specific type of friend recall task. Friend task data includes the task data corresponding to the issued task type, such as the number of friend tasks, the number of completed friend tasks, the task completion time, and the friend return time. The alert period refers to the time from issuing a friend recall task to obtaining the task type and task data. This alert period is determined by domain experts based on past data from issued friend recall tasks and is typically one week.

[0175] Step 70: Analyze the upward trend of friend return time based on the early warning cycle.

[0176] An upward trend refers to the change in the duration of friend return times over time within the warning period. This is analyzed by constructing a friend return time curve with return time on the horizontal axis and daily task duration on the vertical axis, and calculating the overall slope of the curve.

[0177] Step 71: When the upward trend is 0, no action is taken.

[0178] When the upward trend is 0, it means that the published friend recall task has not achieved the effect of recalling friends, so no action should be taken.

[0179] Step 72: When the upward trend is greater than 0 and the friend return time is still less than the preset full return threshold, calculate and analyze the friend churn task completion rate based on the number of friend task completions and the number of friend task completions.

[0180] The full regression threshold refers to the time threshold at which a friend's return time reaches the standard for a complete return. The preset method is similar to the warning cycle. The number of completed friend tasks refers to the number of tasks completed by friends and users included in the friend task data. The number of friend tasks refers to the number of tasks that friends and users included in the friend task data need to complete. The friend churn task completion rate is the ratio of completed friend churn tasks to all required friend churn tasks. When calculating and analyzing the friend churn task completion rate, the calculation formula is used. ,in The completion rate of the task related to friend attrition. M represents the number of tasks completed by friends.

[0181] When the upward trend is greater than 0 and the friend return time is still less than the preset full return threshold, it indicates that the friend recall task has attracted game users and friends but may not have achieved the effect of full return to the game. At this time, it is necessary to calculate and analyze the completion rate of the friend churn task.

[0182] Step 73: When the completion rate of the friend churn task exceeds the preset friend churn task completion threshold, remove the friend from the number of churned friends in the social circle.

[0183] The friend churn task completion threshold is a value representing a certain success rate in completing friend churn tasks. It is pre-set based on the completion status and experience of previously released friend recall tasks. When the friend churn task completion rate exceeds the threshold, it indicates that the number of friend recall tasks completed by the friend and user has reached a certain level, and they are no longer considered potential churners. Therefore, this friend is removed from the list of churned friends in the social circle.

[0184] Step 74: When the completion rate of the friend churn task is less than the friend churn task completion threshold, replace the friend recall task from the preset task list.

[0185] The task list is a database used to store, match, and replace various friend reactivation tasks. The task list is manually entered and stored, and its difficulty is adjusted based on completion status. When a task fails to meet its target, the difficulty level can be adjusted accordingly. If the friend churn task completion rate is below the friend churn task completion threshold, it indicates that the number of friend reactivation tasks completed by friends and users has not reached a sufficient level, and the friend reactivation task does not match the shared gaming experience of friends and users. Therefore, the friend reactivation task is replaced from the task list. Task matching and replacement are based on the currently completed reactivation task type, selecting similar and lower-difficulty reactivation tasks from the task list to replace existing friend reactivation tasks.

[0186] Step 75: When the upward trend is greater than 0 and the friend's return time is greater than the full return threshold, remove the friend from the number of friends lost in the social circle.

[0187] If the upward trend is greater than 0 and the friend's return time is greater than the preset full return threshold, it means that the friend recall task is attractive to game users and friends and has achieved the effect of full return to the game. In this case, the friend is removed from the number of friends lost in the social circle.

[0188] This also includes a method for changing friend recall tasks from a preset task list:

[0189] Step 8: Define friend recall tasks with a friend churn task completion rate lower than the friend churn task completion threshold as failed recall tasks.

[0190] Step 80: Determine the corresponding task data threshold based on the recall task type.

[0191] Task data thresholds refer to the data thresholds corresponding to various data items for each type of recall task, including friend task time thresholds and friend task completion thresholds. The corresponding task data thresholds are determined by searching for friend task times and the number of completed friend tasks based on the friend recall task type.

[0192] Step 81: When the friend task time exceeds the friend task time threshold and the number of friend tasks completed exceeds the friend task completion threshold, a high-difficulty task is matched in the recall task library based on the recall task type, and the friend recall task is replaced.

[0193] If both the friend's task duration and the number of completed tasks exceed the friend's task completion threshold, it indicates that the number of tasks completed by both the friend and the user within the same timeframe is higher than normal. This suggests that the difficulty of the friend recall task is too low for both the friend and the user. In this case, a higher-difficulty task is matched from the recall task library to replace the friend recall task. The task matching and replacement process is described in step 84.

[0194] Step 82: When the friend task time exceeds the friend task time threshold and the number of friend tasks completed is less than the friend task completion threshold, a low-difficulty task is matched in the recall task library based on the recall task type, and the friend recall task is replaced.

[0195] If the friend's task duration exceeds the friend's task duration threshold and the number of completed friend tasks is less than the friend's task completion threshold, it indicates that the number of tasks completed by the friend and user within the same time frame is less than normal. This suggests that the difficulty of the friend recall task is too high for the friend and user. In this case, a lower-difficulty task is matched from the recall task library to replace the friend recall task. The task matching and replacement process is as described in step 84.

[0196] Based on the same inventive concept, embodiments of the present invention provide a game user churn early warning system.

[0197] Reference Figure 5 A game user churn early warning system includes:

[0198] The acquisition module is used to acquire game data from friends, game chat data, data on shared tasks with non-friends, expected task release information, expected release time, historical shared task types, historical shared task data, historical login information, and friend task data.

[0199] The memory is used to store the program for a game user churn early warning method as described above.

[0200] The processor loads and executes programs from memory.

[0201] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0202] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for early warning of user churn in games, characterized in that, include: Step 1: Obtain the game data of the game user's friends, which includes the interaction information between the game user and their friends and the friend's personal login information; Step 10: Extract game data from friends based on mutual interaction information to form a social circle set; Step 11: Filter the number and type of friends lost within the social circle; Step 12: Find the corresponding calculation ratio based on the type of friend churn; Step 13: Calculate the friend churn value based on the number of friends churned according to the type of friend churn and the calculation ratio corresponding to the type of friend churn according to the preset weighted sum formula; Step 14: When the friend churn value exceeds the preset friend churn warning value, mark the game user as a high-risk churn user and output a user churn warning signal; Among them, the methods for predicting friend churn after task release based on friends' game data include: Step 110: Obtain the expected release information and expected release time for the task; Step 111: Based on the social circle set, obtain the historical shared task types, historical shared task data, and historical login information of the user and friends; Step 112: Input the historical common task types, historical common task data and historical login information into the preset standard simulation model for correction to obtain the adapted simulation model, and analyze the expected task release information to obtain the expected task type; Step 113: Input the expected task type into the adaptation simulation model to obtain the expected common task data and expected login information; Step 114: Calculate the attraction value based on the expected release time, expected shared task data, and expected login information, and update the friend churn value according to the attraction value to obtain the expected friend churn value.

2. The game user churn early warning method according to claim 1, characterized in that, Specific methods for extracting game data from friends based on mutual interaction information to form social circles include: Step 2: Obtain shared play information between game users and their friends based on social circles. The shared play information includes shared play data types and shared play data values. Step 20: Find the corresponding intimacy threshold based on the data types of games played together; Step 21: If there is at least one shared play data type with a shared play data value greater than the intimacy threshold, then extract the friend's game data into the social circle set; Step 22: If the shared play data values ​​corresponding to all shared play data types are less than the intimacy threshold, then the friend's game data will not be included in the social circle set.

3. The game user churn early warning method according to claim 2, characterized in that, Optimization methods based on the number of friends lost and the calculation ratio corresponding to the type of friend churn include: Step 3: Define the shared play data types in the friend game data of the social circle that meet the condition that the shared play data value is greater than the intimacy threshold as feature data types; Step 30: Calculate the difference between the shared play data value and the intimacy threshold corresponding to the feature data type; Step 31: Convert the difference into the increase in the calculated proportion of the friend churn data according to the preset formula for calculating the excess portion; Step 32: Adjust the calculation ratio according to the increase to obtain a higher calculation ratio; Step 33: After updating the high calculation ratio to the calculation ratio corresponding to the friend churn type, calculate the friend churn value according to the preset weighted sum calculation formula.

4. The game user churn early warning method according to claim 1, characterized in that, Also includes: Step 4: Based on the user's game data, find the user's and their friends' game chat data, and obtain specific vocabulary information based on the game chat data; Step 40: Define the game users where the specific word information is located as circle-based early warning users, and the circle-based early warning users are users who have not yet churned; Step 41: Calculate the number and type of churn for users who received early warnings, and update the data to the number and type of friend churn. Step 42: If there are preset high-risk words in the specific word information, then define the users in the circle where the high-risk words are located as high-risk users in the circle. Step 43: Count the number and type of high-risk churn corresponding to high-risk users in the circle, and calculate the high-risk ratio value based on the number of high-risk churns and the number of warned churns according to the preset high-risk ratio calculation formula; Step 44: Convert the high-risk ratio value according to the preset high-risk ratio calculation formula to obtain the increase ratio of the calculated ratio of the number of high-risk losses; Step 45: Adjust the calculation ratio according to the increase ratio to obtain the high-risk calculation ratio and update it to the calculation ratio.

5. The game user churn early warning method according to claim 1, characterized in that, Also includes: Step 5: Obtain shared task data from non-social circles and non-friends; Step 50: Determine the judgment parameters based on non-friend shared task data; Step 51: Filter the common task data type in the friend churn type and the friend task data corresponding to the common task data; Step 52: Determine the corresponding friend judgment threshold based on the shared task data of friends; Step 53: If the judgment parameter is greater than the friend judgment threshold, the non-friend shared task data is regarded as friend game data and extracted into the social circle set.

6. The game user churn early warning method according to claim 5, characterized in that, If the non-friend shared task data is treated as friend game data and extracted into the social circle, the specific method for calculating the friend churn value based on the number of friends churned by the friend churn type and the calculation ratio corresponding to the friend churn type according to the preset weighted sum formula includes: Step 6: Classify the non-friend shared task data into high-efficiency task churn type, low-efficiency task churn type, and composite task churn type according to the corresponding shared task threshold; Step 60: Calculate the excess value of the specific task data corresponding to the high-efficiency task churn type, low-efficiency task churn type, and compound task churn type and the corresponding common task threshold using the difference calculation method; Step 61: Based on the excess value, calculate the specific increase in the calculation ratio for the specific task according to the preset task churn excess value calculation formula; Step 62: Update the calculation ratio based on the specific improvement to obtain the specific task calculation ratio.

7. The game user churn early warning method according to claim 1, characterized in that, Also includes: Step 7: In response to the friend recall task sent to friends who have been lost in the social circle, obtain the recall task type and corresponding friend task data after the preset warning period. The friend task data includes the number of friend tasks, the number of friend tasks completed, the friend task time, and the friend return time, etc. Step 70: Analyze the upward trend of friend return time based on the preset early warning period; Step 71: When the upward trend is 0, no action is taken; Step 72: When the upward trend is greater than 0 and the friend return time is still less than the preset full return threshold, calculate and analyze the friend churn task completion rate based on the number of friend task completions and the number of friend task completions. Step 73: When the completion rate of the friend churn task is greater than the preset friend churn task completion threshold, remove the friend from the number of friends churned in the social circle. Step 74: When the completion rate of the friend churn task is less than the friend churn task completion threshold, replace the friend recall task from the preset task list. Step 75: When the upward trend is greater than 0 and the friend's return time is greater than the full return threshold, remove the friend from the number of friends lost in the social circle.

8. A method for early warning of game user churn according to claim 7, characterized in that, Methods for changing friend recall tasks from the preset task list include: Step 8: Define friend recall tasks with a friend churn task completion rate lower than the friend churn task completion threshold as failed recall tasks; Step 80: Determine the corresponding task data threshold based on the recall task type. The task data threshold includes the friend task time threshold and the friend task completion threshold. Step 81: When the friend task time exceeds the friend task time threshold and the number of friend tasks completed exceeds the friend task completion threshold, a high-difficulty task is matched in the recall task library based on the recall task type, and the friend recall task is replaced. Step 82: When the friend task time exceeds the friend task time threshold and the number of friend tasks completed is less than the friend task completion threshold, a low-difficulty task is matched in the recall task library based on the recall task type, and the friend recall task is replaced.

9. A game user churn early warning system, characterized in that, include: The acquisition module is used to acquire game data from friends, game chat data, shared task data from non-friends, expected task release information, expected release time, historical shared task types, historical shared task data, historical login information, and friend task data. A memory for storing a program for a game user churn early warning method as described in any one of claims 1 to 8; The processor loads and executes programs from memory.

Images