An interactive data real-time collection and analysis method for a smart classroom

Abstract

The application relates to the technical field of data processing, and discloses an interactive data real-time collection and analysis method for a smart classroom, which comprises the following steps: analyzing multi-source interactive data streams, extracting basic interactive tuple structures and distributing the basic interactive tuple structures to corresponding time slots; calculating a global safe aggregation water line according to a global maximum event timestamp and a maximum tolerance delay parameter, and outputting complete time window data blocks; calculating information entropy values of teaching behavior events, performing hysteresis interval determination in combination with historical states and high-low dispersion threshold values, and determining storage routing states; dynamically generating a compact bitmap structure or a dynamic inverted index linked list according to the states; responding to instructions to perform cross matching and time sequence constraint checking, and outputting analysis results. The application solves the problems of out-of-order and delay of data transmission, adaptively matches a bottom-layer storage structure to reduce operation internal consumption, reduces invalid memory addressing through pre-filtering, improves retrieval efficiency, and quickly and accurately outputs an academic situation analysis result.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method for real-time acquisition and analysis of interactive data for smart classrooms. Background Technology

[0002] Real-time collection and analysis of interactive data in smart classrooms is a technical means to capture, integrate, and mine multi-source interactive behavioral data generated during the teaching process. In digital education scenarios, student terminal devices continuously generate interactive actions such as answering questions, asking questions, and discussing, forming a massive multi-source interactive data stream. Real-time collection and analysis of these interactive data streams is the basis for evaluating the teaching status and providing decision support.

[0003] The existing interactive data processing mechanism is based on the edge gateway to aggregate behavioral messages uploaded by various terminal devices. After passing through, the messages are sent to the cloud support platform. The cloud system extracts basic information such as timestamps, student identifiers and event types from the messages and performs time-series arrangement and aggregation in memory. Then, according to the set retrieval rules and mining instructions, the system traverses the storage structure in memory to match and statistically analyze the distribution patterns and state characteristics of teaching behaviors, and finally outputs the corresponding learning analysis results to the teaching administrators.

[0004] Currently, network congestion causes out-of-order and delayed arrival of interactive data streams. Existing processing logic lacks a dynamic fault-tolerant truncation mechanism, and late data directly disrupts the consistency of time window calculations, reducing the accuracy of underlying data collection and processing. At the same time, student behavior in the classroom presents either a unified group state or a fragmented individual state. The existing system fails to absorb numerical fluctuations during state transitions using hysteresis judgments and cannot adaptively match the underlying storage structure to different teaching behavior characteristics. This results in the underlying storage structure frequently triggering destruction and reconstruction instructions due to minor fluctuations, increasing server computational overhead and wasting memory space. Furthermore, the existing system lacks pre-filtering conditions when responding to external learning information mining instructions. When performing cross-matching and retrieval, it generates a large number of invalid memory addressing and comparison operations. Traversal operators blindly perform deep extraction when reading data nodes, slowing down data retrieval efficiency and delaying the output speed of analysis results.

[0005] Therefore, the purpose of this invention is to provide a method for real-time acquisition and analysis of interactive data for smart classrooms, in order to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a real-time interactive data acquisition and analysis method for smart classrooms. This method solves the problems of late data disrupting the consistency of time-series calculations, frequent rebuilding of storage structures increasing server computational overhead, and the lack of pre-filtering conditions slowing down data retrieval efficiency in existing interactive data processing mechanisms.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for real-time collection and analysis of interactive data in smart classrooms includes:

[0009] Receive the multi-source interactive data stream transmitted through the edge gateway, and parse and extract the basic interactive tuple structure containing teaching behavior event types, timestamps and student identifiers;

[0010] The basic interactive tuple structure is assigned to the corresponding time slot according to the timestamp;

[0011] The global security aggregation water level is calculated based on the global maximum event timestamp of the multi-source interactive data stream and the preset maximum tolerance delay parameter. The global security aggregation water level is compared with the time boundary of the time slot, and the time window data block of the complete state is output.

[0012] Calculate the probability of occurrence of the teaching behavior event type within the time window data block, and calculate the corresponding information entropy value.

[0013] Obtain historical routing status, and combine the historical routing status, preset high discrete threshold and low discrete threshold and information entropy value to execute hysteresis interval determination logic to determine the current data block storage routing status;

[0014] Based on the current data block storage routing status, a compact bitmap structure or a dynamic inverted index linked list is generated. In response to the learning situation mining command, cross-matching and time-series constraint verification are performed, and the target learning situation analysis results are output.

[0015] Preferably, the step of parsing and extracting the basic interactive tuple structure containing teaching behavior event types, timestamps, and student identifiers from the multi-source interactive data stream transmitted by the receiving edge gateway specifically includes:

[0016] Upon receiving discrete behavior messages, the edge gateway performs protocol conversion and connection multiplexing operations on the aggregated discrete behavior messages, and encapsulates and transparently transmits them as the multi-source interactive data stream;

[0017] Perform a deserialization operation on the multi-source interactive data stream to restore it into independent data frame objects;

[0018] The data frame object is subjected to field identification and extraction, and the extracted fields are instantiated into the basic interactive tuple structure, which includes the timestamp, the student identifier and the teaching behavior event type.

[0019] The above steps perform deserialization and field extraction of network flow to memory objects, transforming the data uploaded by the underlying physical device into a standardized object carrier, providing a unified operation format for subsequent system-executed time-dimensional alignment calculations.

[0020] Preferably, the step of allocating the basic interactive tuple structure to the corresponding time slot according to the timestamp specifically includes:

[0021] Construct a circular buffer based on a one-dimensional array, divide the time slots into multiple time slots according to a preset fixed time granularity, and sequentially and cyclically map the multiple time slots to the physical index of the one-dimensional array of the circular buffer;

[0022] The absolute time slot number is obtained by dividing the timestamp by the fixed time granularity. The actual physical offset address is calculated by performing a modulo operation between the absolute time slot number and the length of the circular buffer.

[0023] Based on the actual physical offset address, locate the target position within the circular buffer, and append the basic interactive tuple structure to the memory data linked list pointed to by the target position to complete the allocation to the corresponding time slot.

[0024] By constructing an array-based circular buffer and combining it with timestamp modulo operations to calculate the actual physical address, the system directly performs allocation and write operations to the target memory block, eliminating the problem of out-of-order and misaligned packets caused by network transmission.

[0025] Preferably, the step of calculating a global security aggregation watermark based on the global maximum event timestamp of the multi-source interactive data stream and a preset maximum tolerance delay parameter, comparing the global security aggregation watermark with the time boundary of the time slot, and outputting a complete state time window data block specifically includes:

[0026] Real-time tracking and recording of the global maximum event timestamp in the multi-source interactive data stream; and calculation of the global safety aggregation water level based on the subtraction of the global maximum event timestamp and the preset maximum tolerance delay parameter.

[0027] When the value of the global security aggregation water level increases monotonically over time and is greater than or equal to the upper limit boundary time of a certain time slot, all the basic interactive tuple structures contained in the time slot that has been collected are extracted and encapsulated and output as the time window data block of the complete state.

[0028] The above steps establish fault-tolerant truncation logic for late data, calculate the global water level and combine it with boundary time limit window output conditions to allow data to tolerate delays within a set time span, while ensuring that the data submitted to the downstream calculation maintains a unidirectional increasing time sequence.

[0029] Preferably, the step of statistically analyzing the occurrence probability of the teaching behavior event type within the time window data block and calculating the corresponding information entropy value specifically includes:

[0030] The occurrence frequency of each non-repeating teaching behavior event type in the time window data block is counted, and the occurrence frequency is divided by the total number of basic interactive tuple structures contained in the time window data block to obtain the occurrence probability of each teaching behavior event type.

[0031] Based on the occurrence probabilities of each item, the corresponding information entropy values ​​are calculated using a metric model in information theory.

[0032] By using the probability of event occurrence to derive the corresponding information entropy value, the state performance of classroom operations can be quantitatively defined. Based on the derived value, the fragmentation tendency of individual behavior or the level of consistency of group operations can be measured, and a data decision-making benchmark for structural allocation mechanism can be established.

[0033] Preferably, the step of obtaining historical routing status, and combining the historical routing status, preset high and low discrete thresholds, and the information entropy value to perform hysteresis interval determination logic to determine the current data block storage routing status specifically includes:

[0034] Extract the historical routing status from the previous calculation cycle; if the information entropy value is greater than or equal to the high discrete threshold, assign the current data block storage routing status value to 1;

[0035] If the information entropy value is less than or equal to the low discrete threshold, the current data block storage routing state is assigned a value of 0;

[0036] If the information entropy value is strictly greater than the low discrete threshold and strictly less than the high discrete threshold, and is in the hysteresis dead zone, the state transition operation is not performed, and the extracted historical routing state value is assigned to the current data block to store the routing state.

[0037] By setting a dual-threshold to generate a numerical dead zone, the fluctuations in information entropy calculations during the boundary of business states are absorbed, avoiding the triggering of recycling and reconstruction instructions of the underlying storage structure due to minor data jitter at a single threshold judgment point, thereby reducing the internal consumption of system computing resources.

[0038] Preferably, the step of generating a compact bitmap structure or a dynamic inverted index linked list based on the current data block storage routing state specifically includes:

[0039] Grouping the basic interaction tuple structures included in the time window data block in the memory according to the student identifiers, sorting the groups in chronological order according to the timestamps, extracting and splicing the teaching behavior event types to generate an ordered behavior sequence, and performing hash calculation on the ordered behavior sequence to generate a globally unique composite event key value;

[0040] If the value of the current data block storage routing status is 0, instantiate a bitmap memory block with the composite event key value as the retrieval key, query the preloaded student roster mapping table, convert the student identifier into the corresponding static continuous number, use a bit operation instruction to set the value of the bit pointed to by the static continuous number as the displacement parameter in the bitmap memory block to 1, and complete the construction of the compact bitmap structure;

[0041] If the value of the current data block storage routing status is 1, instantiate the dynamic inverted index linked list with the composite event key value as the retrieval key, and calculate the difference between the timestamp and the lower boundary time of the time window data block to generate a time offset value;

[0042] Instantiate an index data node in the memory, encapsulate and store the student identifier and the time offset value inside the index data node, append and insert the index data node to the tail of the corresponding dynamic inverted index linked list, and complete the construction of the dynamic inverted index linked list.

[0043] The system determines the underlying retrieval container in response to the quantitative status, performs bitmap compression operation to implement status marking in the group consistent state, instantiates the inverted linked list to maintain fine-grained time series information in the individual behavior fragmented state, and improves the adaptability of the data structure to business characteristics.

[0044] Preferably, after the step of outputting the complete time window data block, a step of asynchronous data backup is further included, which specifically includes:

[0045] Synchronously acquiring the basic interaction tuple structures included in the time window data block, and mapping the physical memory address space of the basic interaction tuple structures to the addressing area of the message buffer queue matched with an independent asynchronous I / O thread pool;

[0046] Polling the status of the message buffer queue, when the total number of the basic interaction tuple structures reaches the batch write条数 threshold or the time interval reaches the time refresh threshold, converting the basic interaction tuple structures into binary data streams and encapsulating them into batch insertion request messages;

[0047] Based on the timestamp field recorded in the basic interactive tuple structure, the corresponding date and time slice parameters are calculated to determine the physical addressing path of the target disk storage node in the underlying cloud database array, and the batch insert request message is written into the target disk storage node of the underlying cloud database array.

[0048] A bypass mechanism is used to create isolated memory mappings and independent access thread groups to implement data backup, thus eliminating the blocking effect of the underlying physical hard disk persistence on the main memory computing program and ensuring that all interactive data is completely written to the disk.

[0049] Preferably, the steps of responding to the learning situation mining command by performing cross-matching and temporal constraint verification, and outputting the target learning situation analysis results specifically include:

[0050] The learning information mining instructions are parsed into a set of target composite events. Based on the preset maintained metadata registry, the underlying storage structure type mapped and bound to each target composite event in the set of target composite events is determined.

[0051] If the target composite event set contains target composite events mapped to the compact bitmap structure, perform parallel bitwise AND operations on the extracted multiple compact bitmap structures, clear the bits that do not satisfy the cross behavior characteristics, and encapsulate the output bit array object into a candidate mask set.

[0052] If the target composite event set contains target composite events mapped to the dynamic inverted index linked list, the candidate mask set is extracted as a memory-level pre-filter condition and pushed down to the traversal operator of the dynamic inverted index linked list.

[0053] The system extracts the structure type from the parsed operation instruction lookup table, calls the vector instructions of the underlying hardware to perform bitwise operations on the low discrete bitmap data to filter out irrelevant behavior identifiers, and inputs the generated bitmask into the inverted index structure to perform a joint query operation.

[0054] Preferably, after the step of extracting the candidate mask set as a memory-level pre-filtering condition and pushing it down to the traversal operator of the dynamic inverted index linked list, the method further includes:

[0055] The traversal operator first extracts the student identifier encapsulated in the index data node and reverses it into the corresponding static continuous number. The static continuous number is then used to address the corresponding target bit in the candidate mask set.

[0056] If the value of the target bit is 0, the reading of the remaining load data in the current index data node is blocked, and the execution pointer is modified to the next index data node of the dynamic inverted index linked list;

[0057] If the value of the target bit is 1, extract the time offset value encapsulated in the current index data node, compare the time offset value with the start threshold and end threshold of the time interval carried by the learning situation mining instruction, collect matching node records that meet the threshold range and assemble them into a result list object, and output the target learning situation analysis result.

[0058] Before reading the deep load data inside the structure, the operator verifies the numerical markers pointed to by the candidate mask set, terminates the remaining memory access requests for irrelevant objects, and reduces the amount of timing judgment operations for non-target records to improve the data parsing speed of the comprehensive results.

[0059] This invention provides a method for real-time acquisition and analysis of interactive data in smart classrooms. It offers the following advantages:

[0060] 1. This invention solves the problems of out-of-order and delay issues in the network transmission of interactive data streams within the physical space of a smart classroom by setting up a circular buffer and time slots, and calculating a global safety aggregation water level based on the global maximum event timestamp and maximum tolerance delay parameter of the multi-source interactive data stream. By dynamically comparing the global safety aggregation water level with the time boundary of the time slot, it can solve the problems of out-of-order and delay issues in the network transmission of interactive data streams within the physical space of a smart classroom. It performs dynamic fault-tolerant truncation on late data that seriously exceeds the tolerance delay range, thereby preventing extreme network congestion from destroying the calculation consistency of the time window, ensuring that the system can stably output complete and time-ordered time window data blocks, and improving the accuracy of the underlying interactive data acquisition and processing.

[0061] 2. This invention sets up steps for statistically analyzing the frequency of teaching behavior event types and calculating corresponding information entropy values. It combines historical routing status, high discrete thresholds, and low discrete thresholds to execute hysteresis interval judgment logic. Then, based on the judged routing status, it dynamically generates a compact bitmap structure or a dynamic inverted index linked list. This can quantitatively identify whether student behavior in the smart classroom is in a unified group state or an individual fragmented state. It uses the hysteresis dead zone interval to absorb the numerical fluctuations during the transition period of classroom teaching status, avoiding frequent destruction and reconstruction instructions of the underlying storage structure due to small fluctuations. At the same time, it adaptively matches the most suitable underlying storage structure for different teaching behavior characteristics, maximizing the use of memory space resources and reducing the computational internal consumption of the server.

[0062] 3. This invention sets up logic for cross-matching and timing constraint verification in response to learning information mining instructions. It performs bitwise AND operations on a compact bitmap structure to generate a candidate mask set, and pushes the candidate mask set as a memory-level pre-filter condition into the traversal operator of the dynamic inverted index linked list. This can pre-screen static consecutive numbers that do not meet the cross-matching behavior characteristics through bitwise operations. When traversing index data nodes, if the target bit address does not meet the filtering condition, it directly blocks the deep reading of the remaining load data in the current node. This reduces invalid memory addressing and comparison operations, improves data retrieval efficiency, and ensures that the system can respond quickly and output accurate target learning information analysis results. Attached Figure Description

[0063] Figure 1 This is a flowchart illustrating the overall method of the present invention;

[0064] Figure 2 This is a system architecture diagram of the present invention;

[0065] Figure 3 This is a flowchart of the aggregation module fault-tolerant truncation and output method of the present invention;

[0066] Figure 4 This is a flowchart of the cross-modal joint retrieval method for the analysis module of the present invention;

[0067] Figure 5 This is a line graph illustrating the evolution of information entropy of classroom interactive behavior and the switching of storage routing states, as presented in this invention.

[0068] Figure 6 This is a bar chart comparing the delay of joint retrieval of complex learning situations under different teaching conditions between the present invention and traditional solutions. Detailed Implementation

[0069] The technical solutions in 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.

[0070] See attached document Figure 1 With appendix Figure 2 This invention is based on a real-time interactive data acquisition and analysis method for smart classrooms, including a real-time interactive data acquisition and analysis system for smart classrooms, which includes terminal devices, edge gateways, and a cloud support platform.

[0071] Terminal devices are deployed within the physical space of the smart classroom. They collect interactive behavior information to generate initial data tuples and send these tuples to the edge gateway. The edge gateway and terminal devices are connected via a local area network, aggregating the initial data tuples sent by each terminal device and establishing a unified data transmission channel to forward the uplink data stream to the cloud support platform. The cloud support platform receives the interactive data stream and executes data processing and analysis logic. The cloud support platform integrates access modules, aggregation modules, routing modules, indexing modules, storage modules, and analysis modules.

[0072] See attached document Figure 1 This invention provides a method for real-time collection and analysis of interactive data in smart classrooms, specifically including the following steps:

[0073] S10, the access module receives the multi-source interactive data stream transmitted by the edge gateway, parses and extracts the basic interactive tuple structure, the basic interactive tuple structure records the timestamp, student identifier and teaching behavior event type.

[0074] S20, the aggregation module configures a circular buffer in the memory area, maps the basic interactive tuples according to the timestamp and allocates them to the corresponding time slots;

[0075] S30, the aggregation module calculates the global safe aggregation water level based on the maximum tolerance delay parameter, performs micro-aggregation operation by comparing the water level with the time boundary of the time slot, and outputs the time window data block with complete status.

[0076] S40, the routing module receives the time window data block of the complete status, counts the frequency distribution of various teaching behavior event types in the window, and calculates the corresponding information entropy value;

[0077] S50, the routing module obtains the historical routing status of the previous calculation cycle, and performs hysteresis interval judgment logic in combination with the preset high discrete threshold and low discrete threshold to determine the current data block storage routing status;

[0078] S60, the index module performs asymmetric heterogeneous index tree construction in memory based on the determined storage routing state, generating a compact bitmap structure or a dynamic inverted index linked list;

[0079] S70, the storage module synchronously obtains the original interactive tuple data after micro-aggregation, and writes it to the underlying cloud database array through an asynchronous bypass mechanism to complete the basic data backup;

[0080] S80, the analysis module receives the learning situation mining instruction, parses it into a target composite event set, extracts the candidate mask set and pushes it down to the traversal operator of the dynamic inverted index linked list to perform cross-matching and time-series constraint verification, and outputs the target learning situation analysis results.

[0081] In the specific implementation process, in order to achieve standardized and unified processing of multi-source heterogeneous terminal data, the system first needs to receive and format the discrete raw messages uploaded by the underlying physical devices. Step S10 specifically realizes the access and parsing of multi-source interactive data streams through the following sub-steps:

[0082] S101, the terminal device establishes a local area network communication link with the edge gateway. The terminal device generates a behavior message in response to a button or touch operation and sends the behavior message to the edge gateway. The edge gateway performs protocol conversion and connection multiplexing operations on the aggregated discrete behavior messages, encapsulates the discrete behavior messages into a multi-source interactive data stream, and transmits it transparently to the access module of the cloud support platform. For the specific implementation of the edge gateway performing protocol conversion and multiplexing, those skilled in the art can use existing message queue telemetry transmission protocols or transmission control protocols to establish a data channel. The above network communication mechanism is a well-known technology in this field and will not be described in detail here.

[0083] S102, the access module maintains a data listening state with the edge gateway on the server side, continuously reads the incoming byte stream payload, and performs deserialization operation on the multi-source interactive data stream according to the pre-set data exchange specifications, splitting the continuous data stream and restoring it into independent data frame objects. The multi-source interactive data stream is encapsulated in a lightweight data exchange format during network transmission. The access module identifies and separates the underlying payload data according to the key-value pair mapping relationship defined in the data format.

[0084] S103, the access module performs field identification and extraction on the parsed data frame object, and instantiates it into a basic interactive tuple structure in memory space. The basic interactive tuple structure contains timestamp, student identifier and teaching behavior event type data items.

[0085] In the basic interactive tuple structure, the timestamp records the absolute time when the terminal device captures the operation command, using a millisecond-level integer value calculated from the reference base time; the student identifier is used to uniquely identify the interaction source that triggers the operation globally, and is configured using a fixed-length string format; the teaching behavior event type is used to map specific interactive actions, using an enumerated integer encoding to represent different business instruction categories. After the access module completes the assembly of the basic interactive tuple structure, it passes the basic interactive tuple structure to the internal lock-free queue, providing it to downstream processing nodes for timing calculation.

[0086] After the access module outputs a standardized basic interactive tuple structure, in order to eliminate the out-of-order problem caused by network transmission and to provide a continuous time base for subsequent teaching state machine calculations, the system needs to physically align the discrete data tuples in the time dimension. Step S20 specifically includes the following sub-steps: constructing a memory-based time slot mapping mechanism:

[0087] S201, the aggregation module requests contiguous physical memory space in the server memory area and constructs a circular buffer based on a one-dimensional array. The aggregation module sets the length of the circular buffer according to the system's maximum concurrent processing capacity and tolerable latency range. The aggregation module configures lock-free concurrency control logic in the append operation at the end of the circular buffer. For the specific implementation of lock-free concurrency control, those skilled in the art can use an atomic comparison and exchange instruction mechanism to ensure the atomicity of memory writes. The lock-free concurrency mechanism is a well-known technology in this field.

[0088] S202, the aggregation module discretizes the continuous time axis according to a preset fixed time granularity, dividing it into multiple time slots. The aggregation module sequentially maps the time slots to the physical indices of the one-dimensional array of the circular buffer. The fixed time granularity is configured with millisecond-level parameters to align out-of-order data transmitted over the network within the same window.

[0089] S203, the aggregation module reads the basic interaction tuple structure from the lock-free queue inside the access module, extracts the timestamp recorded in the basic interaction tuple structure, divides the timestamp by a fixed time granularity to obtain the absolute time slot number, and performs a modulo operation between the absolute time slot number and the length of the circular buffer to calculate the actual physical offset address of the basic interaction tuple structure in the circular buffer.

[0090] S204, the aggregation module locates the target position within the circular buffer based on the actual physical offset address. The aggregation module appends the basic interactive tuple structure to the memory data linked list pointed to by the target position, completing the mapping and allocation operation of the basic interactive tuple structure to the time slot, and outputting batch data discretely aggregated by time slot.

[0091] See attached document Figure 3 After the basic interactive tuple structure is mapped to the circular buffer, in order to prevent extreme delay data caused by network congestion from compromising the computational consistency of the time window, the system introduces a dynamic fault-tolerant truncation mechanism for late data. Step S30 specifically calculates the safety waterline and outputs the aggregated data block through the following execution sub-steps:

[0092] S301, during the process of receiving and allocating the basic interactive tuple structure, the aggregation module tracks and records the global maximum event timestamp in the multi-source interactive data stream in real time. Based on the extracted global maximum event timestamp and the preset maximum tolerance delay parameter, the aggregation module calculates the global safe aggregation water level. The specific calculation formula is as follows:

[0093] ;

[0094] In the formula, Represents the overall security water level, measured in milliseconds; Represents the global maximum event timestamp, in milliseconds; This represents the maximum tolerable delay parameter. The maximum tolerable delay parameter defines the time span during which the system allows data packets to arrive later than the normal order. The maximum tolerable delay parameter is a positive integer in milliseconds greater than 0, and its specific value is determined based on network latency test data from the deployment environment.

[0095] In step S302, the aggregation module dynamically compares the calculated global safety aggregation watermark with the time boundaries of each time slot within the circular buffer. Each time slot has a lower and upper boundary time. When the global safety aggregation watermark monotonically increases over time, and its value is greater than or equal to the upper boundary time of a particular time slot, the aggregation module determines that data collection within the corresponding time period for that time slot is complete. The aggregation module extracts all basic interaction tuple structures contained within the completed time slot, encapsulates these structures into a complete time window data block, and releases the memory occupied by the corresponding time slot in the circular buffer.

[0096] S303, when the aggregation module continuously receives new basic interactive tuple structures, it compares the timestamp of the new basic interactive tuple structure with the current global security aggregation level. If the timestamp of the new basic interactive tuple structure is less than the current global security aggregation level, it indicates that the new basic interactive tuple structure is severely late data that exceeds the maximum tolerance delay parameter.

[0097] The aggregation module prevents severely late data from entering the circular buffer to participate in regular time-series aggregation. Through the exception handling link, severely late data is sent directly to the underlying log system for separate disk backup. The exception handling link ensures that the time window data blocks with complete status received by downstream modules maintain a unidirectional increasing order in the time dimension.

[0098] For the time window data block of complete status output by the aggregation module, in order to accurately identify whether the current smart classroom business is in a state of unified group behavior or individual free discussion, the system introduces a measurement model from information theory to quantify the data distribution. Step S40 specifically includes the following sub-steps: calculating the information entropy value of the interactive behavior characteristics:

[0099] S401, the routing module receives the complete state time window data block transmitted by the aggregation module, and the routing module traverses all the basic interactive tuple structures contained in the complete state time window data block to extract all teaching behavior event type data.

[0100] S402, the routing module detects the total number of basic interactive tuple structures contained in the time window data block of the complete state. If the total number of basic interactive tuple structures is 0, the routing module discards the current time window data block of the complete state and waits for the next time window to arrive to prevent the system from generating calculation errors. If the total number of basic interactive tuple structures is greater than 0, the routing module counts the occurrence frequency of each non-repeating teaching behavior event type in the time window data block of the complete state. The routing module divides the occurrence frequency of each teaching behavior event type by the total number of basic interactive tuple structures contained in the time window data block of the complete state to calculate the occurrence probability of each teaching behavior event type.

[0101] S403, the routing module calculates the corresponding information entropy value based on the calculated probability of occurrence of various teaching behavior event types. The specific calculation formula is as follows:

[0102] ;

[0103] In the formula, Represents the value of information entropy; The total number of unique teaching behavior event types contained within a time window data block representing a complete state; Index number representing the type of teaching behavior event; Representing the The probability of occurrence of various teaching behavior event types; Representing the The probability of occurrence of various teaching behavior event types is The logarithm with base 0.

[0104] S404, the routing module quantifies the discrete characteristics of student group behavior in the smart classroom by using information entropy values. The lower the information entropy value, the more homogeneous the teaching behavior event types are in the data block of the time window of the complete state. At this time, the routing module determines that the classroom business is in a state of group consistent behavior, such as the whole class signing in or the whole class submitting option operations synchronously.

[0105] The higher the information entropy value, the more heterogeneous the teaching behavior event types are in the time window data block of the complete state. At this time, the routing module determines that the classroom business is in an individual fragmented behavior state, such as students thinking independently, frequently canceling and modifying answers. The routing module records the calculated information entropy value in the cache as the basis for subsequent storage structure switching.

[0106] After obtaining the information entropy value used to quantify classroom behavior characteristics, in order to avoid the system frequently triggering the destruction and reconstruction of the underlying storage structure due to small fluctuations in the value near the state boundary, the system is configured with an anti-jitter control strategy. Step S50 specifically determines the storage routing status of the current data block through the following execution sub-steps:

[0107] S501, the routing module reads and loads pre-set high and low discrete thresholds from system memory. The high discrete threshold is greater than the low discrete threshold. The specific values ​​of the high and low discrete thresholds are determined based on the total number of non-repeating teaching behavior event types within the time window. Since the theoretical maximum boundary of the information entropy value is the logarithm of this total number to base 2, the routing module sets the high discrete threshold to 80% of the theoretical maximum boundary and the low discrete threshold to 20% of the theoretical maximum boundary. The routing module retrieves the historical routing state from the memory register of the previous calculation cycle. The value of the historical routing state uses integer binary configuration, with a value of 1 indicating that the previous cycle used a high discrete storage scheme and a value of 0 indicating that the previous cycle used a low discrete storage scheme.

[0108] S502, the routing module introduces hysteresis interval determination logic. Combining the information entropy value, high discrete threshold, low discrete threshold, and historical routing status calculated in step S40, it determines the current data block storage routing status. The specific control formula for the hysteresis interval determination logic is as follows:

[0109] ;

[0110] In the formula, This represents the current data block storage routing status; Represents the value of information entropy; Represents a highly discrete threshold; Represents a low discrete threshold; Represents the historical routing status.

[0111] S503, the routing module performs a branch judgment operation based on the control formula. If the information entropy value is greater than or equal to the high discrete threshold, the routing module determines that the student behavior in the current classroom is in a divergent state and assigns the current data block storage routing state value to 1; if the information entropy value is less than or equal to the low discrete threshold, the routing module determines that the student behavior in the current classroom is in a uniform state and assigns the current data block storage routing state value to 0.

[0112] S504. If the information entropy value is strictly greater than the low discrete threshold and strictly less than the high discrete threshold, it indicates that the information entropy value is in the hysteresis dead zone. When it is in the hysteresis dead zone, the routing module directly assigns the value of the historical routing state to the current data block storage routing state without performing a state transition operation. The routing module uses the hysteresis dead zone to absorb the numerical fluctuations during the transition of the classroom teaching state, preventing the underlying storage engine from frequently triggering resource reclamation and structure reconstruction instructions when the information entropy value fluctuates slightly near a single judgment critical point, thereby reducing the computational internal consumption of the system. After the routing module completes the judgment, it will pass the determined current data block storage routing state to the downstream nodes.

[0113] Based on the storage routing status determined by the routing module, in order to maximize the utilization of server memory space resources and improve retrieval efficiency for different behavioral characteristics, the system needs to load the interaction data into the matching underlying retrieval data structure. Step S60 specifically includes the following sub-steps: constructing an asymmetric heterogeneous memory index tree:

[0114] S601, the index module receives the complete state time window data block and the current data block storage routing state passed through by the routing module. The index module groups the basic interaction tuple structure contained in the complete state time window data block into memory groups according to student identifiers. In the memory group corresponding to each student identifier, the index module sorts the basic interaction tuple structure according to the order of timestamps, extracts the teaching behavior event types in sequence, and splices them to generate an ordered behavior sequence.

[0115] The index module performs hash calculations on ordered behavior sequences using a preset hash mapping algorithm to generate globally unique composite event keys. These composite event keys serve as the retrieval entry point for the underlying storage structure. The preset hash mapping algorithm is implemented using the secure hash algorithm SHA-256.

[0116] S602, the index module reads the integer value of the current data block storage routing status and performs an asymmetric heterogeneous index tree construction operation based on the integer value. If the value of the current data block storage routing status is 0, the index module determines that the current system is in a low discrete feature state and triggers the construction link of the compact bitmap structure.

[0117] In S603, in the construction chain of the compact bitmap structure, the index module instantiates a bitmap memory block in memory with the calculated composite event key value as the retrieval key. The index module extracts the student identifiers recorded in the basic interactive tuple structure, queries the student roster mapping table pre-loaded by the system, converts the student identifiers into corresponding static consecutive numbers, and uses the static consecutive numbers as the displacement parameters of the bitmap memory block. The index module uses bitwise operation instructions to set the bit value pointed to by the displacement parameters in the bitmap memory block to 1, thus completing the construction of the compact bitmap structure.

[0118] S604, if the value of the current data block storage routing status is 1, the index module determines that the current system is in a highly discrete characteristic state and triggers the construction link of the dynamic inverted index linked list.

[0119] In S605, during the construction of the dynamic inverted index linked list, the index module instantiates the dynamic inverted index linked list in memory using the composite event key as the retrieval key. The index module extracts the timestamps from the records of the basic interaction tuple structure, calculates the difference between the timestamp and the lower boundary time of the time window data block in the current complete state, and generates a time offset value. The index module instantiates index data nodes in memory, encapsulates the student identifier and the time offset value within the index data node, and appends the encapsulated index data node to the tail of the corresponding dynamic inverted index linked list, completing the construction of the dynamic inverted index linked list.

[0120] While the system allocates computing resources to the construction of the real-time in-memory index, in order to meet the disaster recovery and historical traceability requirements of industrial-grade data services, the system must ensure that all original data is completely persisted to disk. Step S70 specifically performs asynchronous bypass backup of the data without blocking the main computing link through the following execution sub-steps:

[0121] In S701, when the aggregation module outputs a complete time window data block, the storage module synchronously obtains the original interactive tuple data contained in the time window data block. The storage module configures an independent asynchronous I / O thread pool and a corresponding message buffer queue at the server-side operating system level. The storage module calls the operating system's memory mapping instruction to map the physical memory address space containing the original interactive tuple data to the addressing area of ​​the message buffer queue. After completing the memory address mapping, the system's main computing thread executes a return instruction and processes subsequent data block operations. The asynchronous I / O thread pool takes over the processing flow in the message buffer queue. The storage module establishes an asynchronous bypass mechanism by allocating an independent thread group and an isolated memory operation area.

[0122] In the S702, worker threads in the asynchronous I / O thread pool poll the message buffer queue at fixed intervals. The storage module is pre-configured with batch write thresholds and time refresh thresholds. The batch write threshold is a positive integer, configurable based on server memory capacity, ranging from 1000 to 5000 records. The time refresh threshold is a millisecond-level parameter, configurable based on the business's tolerance for data persistence latency, ranging from 1000ms to 3000ms. When the total number of raw interactive tuples stored in the message buffer queue reaches the batch write threshold, or when the time interval from the last write command to the current time refresh threshold reaches the time refresh threshold, the worker thread performs an extraction operation. The worker thread extracts the raw interactive tuple data from the message buffer queue, converts it into a binary data stream using a protocol buffer data serialization mechanism, and encapsulates it into a batch insert request message. The storage module uses the long connection channel maintained by the underlying Transmission Control Protocol (TCP) to send the batch insert request message to the underlying cloud database array.

[0123] S703: The underlying cloud database array receives batch insert request messages, executes deserialization instructions to extract the original interactive tuple data inside the message payload, reads the timestamp field of each original interactive tuple data record, calculates the corresponding date and time slice parameters based on the timestamp field, and determines the physical addressing path of the target disk storage node based on the date and time slice parameters.

[0124] The underlying cloud database array writes the original interactive tuple data to the designated target disk storage node to complete the basic data backup operation. For the specific implementation of distributed sharding and persistence of the underlying cloud database array, those skilled in the art can use a document-oriented distributed database or a columnar time-series database to build the underlying storage cluster.

[0125] See attached document Figure 4 Once the aforementioned heterogeneous underlying storage structure is completed and resides in the memory environment, in order to respond to the real-time joint analysis requirements of external business systems for complex classroom teaching scenarios, the system is configured with a cross-modal joint retrieval engine. Step S80 specifically includes the following execution sub-steps: execution instruction pushdown and cross-matching logic:

[0126] S801, the analysis module receives learning information mining instructions from external business systems. The analysis module calls the built-in lexical and grammatical parsing engine to perform text segmentation on the learning information mining instructions. The analysis module generates an abstract syntax tree based on the preset syntax rule library, traverses the operator nodes and parameter nodes in the abstract syntax tree, and converts the learning information mining instructions into a set of underlying target composite events. The analysis module reads the preset metadata registry in the system memory. The metadata registry uses an in-memory hash table data structure to store the mapping relationship between composite event keys and storage engine types. Based on the mapping relationship, the analysis module determines the underlying storage structure type currently mapped to each target composite event in the target composite event set. The underlying storage structure type is divided into a compact bitmap structure and a dynamic inverted index linked list.

[0127] S802, if the target composite event set contains target composite events mapped to a compact bitmap structure, the analysis module extracts the corresponding bound compact bitmap structure from the memory space, and the analysis module calls the single instruction stream multiple data stream hardware instruction set provided by the underlying central processing unit, specifically using the advanced vector extended instruction set to operate memory.

[0128] The analysis module performs parallel bitwise AND operations on the extracted compact bitmap structures at the hardware register level. Through hardware-level bitwise operations, the analysis module clears the bits that do not meet the crossover behavior characteristics and filters out the static consecutive student numbers that have not exhibited the specified group consistency behavior. The analysis module encapsulates the bit array object finally output by the bitwise AND operation into a candidate mask set and stores it in the cache of the central processing unit.

[0129] S803, if the target composite event set contains target composite events mapped to the dynamic inverted index linked list, the analysis module triggers the heterogeneous storage joint query mechanism. The analysis module extracts the candidate mask set as a memory-level pre-filter condition and pushes the candidate mask set down to the traversal operator of the dynamic inverted index linked list. The traversal operator obtains the head memory address pointer of the dynamic inverted index linked list and accesses each index data node in the order of the unidirectional physical address referenced by the pointer. When reading each index data node, the traversal operator prioritizes extracting the student identifier encapsulated inside the index data node.

[0130] In S804, the traversal operator inputs the extracted student identifier into the student roster mapping table, reverses it to the corresponding static consecutive number, and uses the static consecutive number to calculate the byte offset address. It then addresses the corresponding target bit in the candidate mask set in the cache. If the value of the target bit is 0, the traversal operator determines that the current student identifier does not meet the pre-filtering condition. The traversal operator then blocks the memory controller from reading the remaining load data in the current index data node and directly modifies the execution pointer to the next index data node in the dynamic inverted index linked list.

[0131] If the target bit value is 1, the traversal operator triggers a deep data reading operation, extracts the time offset value encapsulated in the current index data node, reads the start and end thresholds of the time interval carried by the learning situation mining instruction, compares the time offset value with the threshold range, executes the timing constraint verification logic, the analysis module collects all matching node records that meet the threshold range, assembles the student identifier and time attribute in the matching node records into a result list object, and outputs the target learning situation analysis results to the upper-level business system.

[0132] Specific application examples:

[0133] To verify the effectiveness of the proposed real-time interactive data acquisition and analysis method for smart classrooms in solving the problems of network transmission disorder, excessive system storage resource consumption due to frequent switching of classroom states, and high latency in complex learning information retrieval in complex high-concurrency teaching scenarios, this embodiment is based on the interactive teaching application scenario of a standard smart classroom (Room-302) in a university, and combined with the attached... Figure 5 and attached Figure 6 The data shown will be explained in detail.

[0134] Appendix Figure 5 and attached Figure 6 The data in this document are all internal calculation data captured in real time by this system, compared with traditional schemes based on fixed time windows and single database indexes.

[0135] In the application scenario of this embodiment, the smart classroom is equipped with 50 student interactive terminals. The system supports a total of 20 types of teaching behavior events. In response to the out-of-order data packets caused by network fluctuations and the alternation of teaching sessions (such as unified tests for the whole class and free group discussions), the system has formulated a dynamic safety water level fault tolerance and heterogeneous storage routing joint debugging scheme based on information entropy.

[0136] During the process of the system receiving multi-source interactive data streams, the fixed time granularity of the aggregation module is set to 500ms, and the maximum tolerable delay parameter is set according to the latency test calibration of the on-site wireless LAN. .

[0137] During the in-class quiz, the network experienced a brief period of congestion.

[0138] The edge gateway transmits data to the access module, and the aggregation module tracks the global maximum event timestamp in the current multi-source interactive data stream in real time. The cloud support platform uses the global security aggregation water level formula. The current global safe aggregate water level is calculated. At this point, the aggregation module compared and found that the upper limit boundary time of a certain time slot in the circular buffer was 1682003000ms.

[0139] because If the value is greater than the upper limit boundary time of the time slot, the aggregation module determines that the data collection corresponding to the time slot has been completed, and then encapsulates its internal basic interaction tuple structure into a complete time window data block for output. At this time, if a basic interaction tuple structure with a timestamp of 1682002000ms is received, because it is less than... The system identifies it as severely late data and bypasses it directly to the log system for separate disk storage, ensuring the unidirectional increasing order of the main window data blocks in the time dimension.

[0140] After the routing module receives the time window data block of the complete status, in order to verify the effectiveness of the system's quantitative classroom business status and anti-jitter control strategy, its internal behavioral characteristics are calculated.

[0141] After the teacher finishes the standardized test, students are given 5 minutes to discuss and revise their answers in groups.

[0142] From the appendix Figure 5 The data details are available. Figure 5 The middle horizontal axis represents the time axis (numbered according to the time window batch), the left axis represents the information entropy value (in bits), and the right axis represents the storage routing status (0 or 1).

[0143] When the time window batch number was 5, the routing module detected a total of 50 basic interactive tuple structures. Statistical analysis revealed 5 distinct types of teaching behavior events. ), where the frequency and probability of occurrence are as follows: Behavior A (modify answer, probability) Behavior B (researching materials, probability) Behavior C (Cancel operation, probability) Behavior D (submitting a request for help, probability) Behavior E (screen annotation, probability) According to the numerical formula for information entropy: The routing module calculates... The system reads a pre-set threshold. Given that the total number of 20 behaviors has a theoretical maximum bound of approximately 4.32 based on logarithmic theory, a high discrete threshold is set. Low discrete threshold Historical routing status Based on the hysteresis interval determination logic control formula: ;

[0144] At this point, the calculation yields... It is in the hysteresis dead zone, combined with the attached Figure 5 The line trend shows that within the time window batch numbers 3 to 7, as the class gradually transitioned from group discussions to unified testing, the information entropy value gradually decreased from 2.80 to 1.60, remaining within the hysteresis dead zone range of 0.86 to 3.45. During this period, the system, based on the formula... The system directly keeps the current data block storage routing state set to 1 until the batch number in the time window reaches 8, at which point the information entropy value drops to 0.70, falling below the low discrete threshold of 0.86 for the first time. Only then does the system switch the storage routing state to 0 according to the control formula. This process uses the hysteresis dead zone to absorb the numerical fluctuations in the information entropy value during the transition period of the teaching state, avoiding the risk of prematurely switching the underlying storage structure, preventing the underlying storage engine from frequently triggering resource reclamation and structure reconstruction instructions near the critical point, and reducing the system's computational internal consumption.

[0145] After the data was stored in the heterogeneous in-memory index tree, in order to solve the retrieval delay problem when external business systems perform cross-learning mining, the underlying cross-modal joint retrieval engine was compared and verified.

[0146] The business system issued an instruction to select a set of students who "implemented unified sign-in during the quiz and modified their answers during the discussion".

[0147] From the appendix Figure 6 As shown in the bar chart, the horizontal axis represents different query concurrency levels (times / second), and the vertical axis represents the average join retrieval latency (ms). In this query, the target composite event set includes the "unified check-in" event mapped to the compact bitmap structure and the "answer modification" event mapped to the dynamic inverted index. The analysis module first calls the advanced vector extension instruction set to perform a bitwise AND operation on the compact bitmap structure, quickly filtering out students who have not checked in and generating a candidate mask set. Subsequently, this mask is pushed down to the dynamic inverted index linked list. Before accessing nodes, the traversal operator uses static consecutive numbers to address the target bit in the mask, directly skipping irrelevant nodes where the target bit is 0, avoiding full table scans and deep readings of invalid data nodes.

[0148] From the appendix Figure 6 The comparative data shows that under the pressure of query concurrency reaching 500 times / second, the average retrieval latency of traditional single database engines increases to more than 1250ms due to the need to perform a large number of table joins and primary key table lookups. However, after introducing a front-end bitmap filtering and dynamic pointer blocking mechanism, the average joint retrieval latency of this system is stabilized at around 180ms.

[0149] Summary of application examples:

[0150] This embodiment demonstrates the effectiveness of the real-time interactive data acquisition and analysis method for smart classrooms. On one hand, it utilizes a waterline formula based on a tolerance delay parameter to achieve safe truncation of out-of-order data and complete window encapsulation, ensuring the timeliness of calculations. On the other hand, it uses information entropy to quantify group behavior and smooths state switching judgments through a hysteresis dead zone formula, avoiding unnecessary memory resource reorganization and waste. When dealing with complex cross-learning mining commands, the analysis module reduces the amount of memory access to invalid data by using a combined pushdown strategy of asymmetric bitmaps and inverted indexes, ensuring the system maintains high-concurrency data writing while ensuring analysis response capabilities.

Claims

1. A method for real-time acquisition and analysis of interactive data for smart classrooms, characterized in that, Includes the following steps: Receive the multi-source interactive data stream transmitted through the edge gateway, and parse and extract the basic interactive tuple structure containing teaching behavior event types, timestamps and student identifiers; The basic interactive tuple structure is assigned to the corresponding time slot according to the timestamp; The global security aggregation water level is calculated based on the global maximum event timestamp of the multi-source interactive data stream and the preset maximum tolerance delay parameter. The global security aggregation water level is compared with the time boundary of the time slot, and the time window data block of the complete state is output. Calculate the probability of occurrence of the teaching behavior event type within the time window data block, and calculate the corresponding information entropy value. Obtain historical routing status, and combine the historical routing status, preset high discrete threshold and low discrete threshold and information entropy value to execute hysteresis interval determination logic to determine the current data block storage routing status; Based on the current data block storage routing status, a compact bitmap structure or a dynamic inverted index linked list is generated. In response to the learning situation mining command, cross-matching and time-series constraint verification are performed, and the target learning situation analysis results are output.

2. The method for real-time acquisition and analysis of interactive data for smart classrooms according to claim 1, characterized in that, The steps of parsing and extracting the basic interactive tuple structure containing teaching behavior event types, timestamps, and student identifiers from the multi-source interactive data stream transmitted by the receiving edge gateway specifically include: Upon receiving discrete behavior messages, the edge gateway performs protocol conversion and connection multiplexing operations on the aggregated discrete behavior messages, and encapsulates and transparently transmits them as the multi-source interactive data stream; Perform a deserialization operation on the multi-source interactive data stream to restore it into independent data frame objects; The data frame object is subjected to field identification and extraction, and the extracted fields are instantiated into the basic interactive tuple structure, which includes the timestamp, the student identifier and the teaching behavior event type.

3. The method for real-time acquisition and analysis of interactive data for smart classrooms according to claim 2, characterized in that, The step of assigning the basic interactive tuple structure to the corresponding time slot according to the timestamp specifically includes: Construct a circular buffer based on a one-dimensional array, divide the time slots into multiple time slots according to a preset fixed time granularity, and sequentially and cyclically map the multiple time slots to the physical index of the one-dimensional array of the circular buffer; The absolute time slot number is obtained by dividing the timestamp by the fixed time granularity. The actual physical offset address is calculated by performing a modulo operation between the absolute time slot number and the length of the circular buffer. Based on the actual physical offset address, locate the target position within the circular buffer, and append the basic interactive tuple structure to the memory data linked list pointed to by the target position to complete the allocation to the corresponding time slot.

4. The method for real-time acquisition and analysis of interactive data for smart classrooms according to claim 3, characterized in that, The steps of calculating the global security aggregation watermark based on the global maximum event timestamp of the multi-source interactive data stream and the preset maximum tolerance delay parameter, comparing the global security aggregation watermark with the time boundary of the time slot, and outputting the complete state time window data block specifically include: Real-time tracking and recording of the global maximum event timestamp in the multi-source interactive data stream; and calculation of the global safety aggregation water level based on the subtraction of the global maximum event timestamp and the preset maximum tolerance delay parameter. When the value of the global security aggregation water level increases monotonically over time and is greater than or equal to the upper limit boundary time of a certain time slot, all the basic interactive tuple structures contained in the time slot that has been collected are extracted and encapsulated and output as the time window data block of the complete state.

5. The method for real-time acquisition and analysis of interactive data for smart classrooms according to claim 4, characterized in that, The steps of statistically analyzing the occurrence probability of the teaching behavior event type within the time window data block and calculating the corresponding information entropy value specifically include: The occurrence frequency of each non-repeating teaching behavior event type in the time window data block is counted, and the occurrence frequency is divided by the total number of basic interactive tuple structures contained in the time window data block to obtain the occurrence probability of each teaching behavior event type. Based on the occurrence probabilities of each item, the corresponding information entropy values ​​are calculated using a metric model in information theory.

6. The method for real-time acquisition and analysis of interactive data for smart classrooms according to claim 5, characterized in that, The step of obtaining historical routing status and, in conjunction with the historical routing status, preset high and low discrete thresholds, and the information entropy value, performing hysteresis interval determination logic to determine the current data block storage routing status specifically includes: Extract the historical routing status from the previous calculation cycle; If the information entropy value is greater than or equal to the high discrete threshold, the current data block storage routing state is assigned a value of 1; If the information entropy value is less than or equal to the low discrete threshold, the current data block storage routing state is assigned a value of 0; If the information entropy value is strictly greater than the low discrete threshold and strictly less than the high discrete threshold, and is in the hysteresis dead zone, the state transition operation is not performed, and the extracted historical routing state value is assigned to the current data block to store the routing state.

7. The method for real-time acquisition and analysis of interactive data for smart classrooms according to claim 6, characterized in that, The steps for generating a compact bitmap structure or a dynamic inverted index linked list based on the current data block storage routing state specifically include: The basic interactive tuple structure contained in the time window data block is grouped in memory according to the student identifier and sorted in the order of the timestamps within the group. The teaching behavior event types are extracted and concatenated to generate an ordered behavior sequence. The ordered behavior sequence is hashed to generate a globally unique composite event key value. If the value of the current data block storage routing status is 0, instantiate a bitmap memory block with the composite event key as the retrieval key, query the pre-loaded student roster mapping table, convert the student identifier into the corresponding static consecutive number, and use bit operation instructions to set the bit value pointed to by the static consecutive number as the displacement parameter in the bitmap memory block to 1, thereby completing the construction of the compact bitmap structure. If the value of the current data block storage routing status is 1, the dynamic inverted index list is instantiated with the composite event key as the retrieval key, and the difference between the timestamp and the lower boundary time of the time window data block is calculated to generate a time offset value. An index data node is instantiated in memory, and the student identifier and the time offset value are encapsulated and stored inside the index data node. The index data node is then appended to the tail of the corresponding dynamic inverted index linked list, thus completing the construction of the dynamic inverted index linked list.

8. The method for real-time acquisition and analysis of interactive data for smart classrooms according to claim 4, characterized in that, After the step of outputting the time window data block in the complete state, the method further comprises a step of asynchronous data backup, which specifically includes: Synchronously acquiring the basic interaction tuple structure included in the time window data block, and mapping the physical memory address space of the basic interaction tuple structure to the addressing area of the message buffer queue corresponding to an independent asynchronous I / O thread pool; Polling the status of the message buffer queue, and when the total number of the basic interaction tuple structures reaches the batch writing number threshold or the time interval reaches the time refresh threshold, converting the basic interaction tuple structures into binary data streams and encapsulating them into a batch insertion request message; Calculating the corresponding date and time slice parameters according to the timestamp field recorded in the basic interaction tuple structure, determining the physical addressing path of the target disk storage node in the underlying cloud database array, and writing the batch insertion request message into the target disk storage node of the underlying cloud database array.

9. A method for real-time acquisition and analysis of interactive data for smart classrooms according to claim 1, characterized in that, The step of responding to the learning situation mining instruction to perform cross-matching and timing constraint verification and output the target learning situation analysis result specifically comprises: Parsing the learning situation mining instruction into a target composite event set, and determining the underlying storage structure type mapped and bound to each target composite event in the target composite event set according to a preset maintained metadata registry; If the target composite event set contains a target composite event mapped to the compact bitmap structure, performing a parallel bitwise AND operation on the plurality of extracted compact bitmap structures, clearing the bits that do not satisfy the cross behavior characteristic to zero, and outputting the bit array object encapsulated as a candidate mask set; If the target composite event set also contains a target composite event mapped to the dynamic inverted index linked list at the same time, extracting the candidate mask set as an in-memory pre-filtering condition and pushing it down to the traversal operator of the dynamic inverted index linked list.

10. A method for real-time acquisition and analysis of interactive data for smart classrooms according to claim 9, characterized in that, After the step of extracting the candidate mask set as the in-memory pre-filtering condition and pushing it down to the traversal operator of the dynamic inverted index linked list, the method further comprises: The traversal operator preferentially extracts the student identifier encapsulated in the index data node and reversely converts it into a corresponding static continuous number, and uses the static continuous number to address the corresponding target bit in the candidate mask set; If the value of the target bit is 0, blocking reading of the remaining payload data in the current index data node, and modifying the execution pointer to the next index data node of the dynamic inverted index linked list; If the value of the target bit is 1, extracting the time offset value encapsulated in the current index data node, comparing the time offset value with the time interval start threshold and time interval end threshold carried by the learning situation mining instruction, collecting matching node records that meet the threshold range, assembling them into a result list object, and outputting the target learning situation analysis result.

Images