Activity type identification method and system based on daily activity rhythm

Abstract

The invention relates to an activity type identification method and system based on a daily activity rhythm, and the method comprises the steps: identifying activity types based on resident trip data according to an activity type judgment condition, and generating a group daily activity rhythm curve based on the resident trip data for each activity type; the activity types comprise living activities, working activities and other activities; taking the activity type as a label of a daily activity rhythm curve of a corresponding group, constructing a training set, and training the initial activity type classification model by using the training set to obtain an activity type classification model; constructing an individual daily activity rhythm curve based on the mobile communication data, and outputting an individual activity type by using an activity type classification model based on the individual daily activity rhythm curve; the system is used for implementing the method. Compared with the prior art, the generalization ability and accuracy of activity type recognition in complex and heterogeneous behavior scenes are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of spatio-temporal behavior analysis and pattern recognition, and in particular, to a daily activity rhythm-based activity type identification method and system. BACKGROUND

[0002] With the diversification of social employment forms, the daily activities of commuting groups have become increasingly heterogeneous and dynamic. Although traditional active data collection methods can conduct surveys for specific groups, they are difficult to conduct large-scale and continuous activity analysis due to high cost, small sample size, and low update frequency. In contrast, mobile communication data has gradually become the core data source for spatio-temporal behavior research due to its high spatio-temporal resolution and continuous observation advantages. However, mobile communication data only records the location information of individuals at a specific time and place, and lacks semantic annotation of activities, making it difficult to be directly used for activity type identification. Currently, activity type identification mostly uses rule-based methods, which, although have strong interpretability, usually rely on two assumptions: one is that activity time is relatively fixed, for example, residential activities usually occur from 8 pm to 8 am the next day, and work activities occur from 9 am to 5 pm; the other is that the activity location is unique, and individuals usually flow between fixed residential and work locations. For example, Chinese patent application CN107529135A provides a user activity type identification method based on intelligent device data, which includes extracting activity start time and activity duration, and obtaining the land use properties of the stopped section according to the interest points of the stopped section; analyzing the intelligent device data of the user's multi-day travel to determine the location of the user's home and / or work place to obtain the two activity types of being at home or at work; and by inputting the activity features into an activity classifier to obtain the corresponding activity types, the method can achieve comprehensive identification of user travel activities, but it has the problems mentioned above, specifically: the identification of work and residential locations relies on fixed spatio-temporal period rules, which cannot adapt to scenarios with ambiguous work time boundaries such as overtime; and the provided residential and work location identification method is based on single location statistical logic, which cannot adapt to multi-location flow scenarios.

[0003] Therefore, how to avoid the use of fixed spatio-temporal rules to cause inaccurate identification of human activity types is a technical problem to be solved. SUMMARY

[0004] The purpose of the present application is to overcome the defects of the prior art and provide a daily activity rhythm-based activity type identification method, which improves the accuracy and robustness of activity type identification by fusing mobile communication data and resident travel data.

[0005] The purpose of the present application can be achieved by the following technical solutions: According to a first aspect of the present application, a daily activity rhythm-based activity type identification method is provided, the method comprising: identifying an activity type according to an activity type determination condition based on resident travel data, and for each activity type, generating a group daily activity rhythm curve based on the resident travel data; the activity type includes residential activity, work activity and other activity; using the activity type as a label corresponding to the group daily activity rhythm curve, constructing a training set, training an initial activity type classification model using the training set, and obtaining an activity type classification model; constructing an individual daily activity rhythm curve based on mobile communication data, and outputting an individual activity type based on the individual daily activity rhythm curve using the activity type classification model.

[0006] As a preferred technical solution, the activity type determination condition comprises: based on the resident travel data, if the resident's first travel starting location on the day is the residence, the activity type of the time period from 00:00 to the first travel starting time on the day is determined as residential activity; if the resident's last travel arrival location on the day is the residence, the time period from the last travel arrival time to 23:59 on the day is determined as residential activity; based on the travel log recorded in the resident travel data, the travel purpose is determined as work activity, for each of the work activities, the starting time is the time of arriving at the workplace, and the ending time is the departure time of the next travel; and if the travel is the last travel on the day, the ending time of the corresponding work activity is recorded as 23:59 on the day; the other activity is all activities other than work activity and residential activity.

[0007] As a preferred technical solution, the method for generating the group daily activity rhythm curve comprises: in the spatial scale, the collection area is divided into a regular network; in the time scale, the whole day is divided into multiple time intervals; for any activity type, based on the resident travel data, the activity location of each individual in the corresponding activity type is obtained, and the activity location is aggregated in time and space, and the number of group activities in each grid and time interval is counted, that is, the group activity intensity, and the group activity intensity sequence is constructed in the time unit of day: , wherein, represents the group activity intensity of participating in the activity type in the time interval in the grid ; represents the total number of time intervals; each sequence recorded in the group activity sequence After normalization, the rhythm curve of the group's daily activities is obtained as follows: , Indicates the time interval Inside, in the grid Participate in the activity type The group's daily activity points.

[0008] As a preferred technical solution, the method for constructing the individual daily activity rhythm curve is as follows: Based on the aforementioned mobile communication data, individual activity points and activity records are extracted, and individual activity locations are identified based on the aforementioned individual activity points. The individual's multi-day activity records in different activity locations are projected onto a standardized timeline and discretized into multiple time slices at the time level; For each individual activity location, the frequency of its occurrence in different time slices is counted. Based on the frequency of occurrence, the cumulative activity intensity across days is calculated, and a cumulative activity intensity sequence across days is constructed for each activity location. , Indicates an individual's activity location Upper The cumulative activity intensity across days in a given time period, Indicates the total number of time slices; The cumulative daily activity intensity of each activity location in the individual's cumulative daily activity intensity sequence is normalized to construct an individual's daily activity rhythm curve. , Indicates an individual's activity location Upper The individual daily activity rhythm curve points of each time slice.

[0009] As a preferred technical solution, the activity type classification model includes an input layer, multiple fully connected hidden layers, and an output layer; wherein, the input of the input layer is an individual's daily activity rhythm curve; the number of neurons in each hidden layer is different; and the output layer converts the output of the last hidden layer into a probability distribution of activity type through a Softmax activation function.

[0010] According to a second aspect of the present invention, an activity type identification system based on daily activity rhythms is provided, comprising: Data Acquisition and Processing Module: This module collects resident travel data and mobile communication data. Based on the resident travel data, it identifies activity types according to activity type determination conditions. For each activity type, it generates a group daily activity rhythm curve based on the resident travel data. It also constructs an individual daily activity rhythm curve based on the mobile communication data. The activity types include residential activities, work activities, and other activities. Classification model training module: This module uses activity type as the label of the daily activity rhythm curve of the corresponding group, constructs a training set, and uses the training set to train an initial activity type classification model to obtain the activity type classification model; The individual activity type generation module outputs the individual activity type based on the individual daily activity rhythm curve and the activity type classification model.

[0011] Compared with the prior art, the present invention has the following beneficial effects: 1) This invention effectively compensates for the shortcomings of traditional rule-based methods in identifying complex behavioral patterns by constructing and utilizing daily activity rhythms as the core representation for activity type inference. Specifically, by performing time alignment and normalization processing on an individual's multi-day mobile communication data, the daily activity rhythms of the individual in different activity locations are constructed. The resulting long-term behavioral characteristics of the individual no longer depend on fixed time windows or single activity locations, thus avoiding the rule failure problem caused by extended working hours or multi-location activities, thereby achieving more accurate identification of human activity types.

[0012] 2) This invention introduces a training set based on the daily activity rhythms of groups constructed from resident travel data to train the activity type classification model. This enables the activity type classification model to learn the mapping relationship between activity rhythms and activity types, transforming activity type identification from rule matching to pattern learning. This allows for the identification of behavioral patterns that are easily misjudged or ignored in traditional rule-based methods, especially for groups with extended working hours, groups with multiple work locations, and work-driven groups with multiple residences. This improves the generalization ability and accuracy of activity type identification in complex and heterogeneous behavioral scenarios. Attached Figure Description

[0013] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0015] To address the problems existing in current technologies, this invention provides a method for activity type identification based on daily activity rhythms. First, a group daily activity rhythm curve is constructed, and a labeled training set is built based on this curve. This training set is then used to train an activity type classification model, enabling the model to learn the mapping relationship between activity rhythms and activity types. Finally, this model is used to identify individual activity types based on individual daily activity rhythm curves. The detailed method is as follows: Figure 1 As shown, it includes: S1. Identify activity types based on resident travel data and activity type determination conditions, and generate a daily activity rhythm curve for each activity type based on resident travel data.

[0016] S11, Activity type determination.

[0017] In this invention, activity types include residential activities, work activities, and other activities, and the criteria for determining activity types are as follows: Based on residents' travel data, if a resident's first trip of the day starts at their place of residence, then the activity type from 00:00 to the start time of the first trip is determined to be a residential activity; if a resident's last trip of the day arrives at their place of residence, then the time from the arrival time of the last trip to 23:59 is determined to be a residential activity; the above two time periods together constitute the individual's residential activity time period for the day.

[0018] Work activities are determined based on the travel logs recorded in the residents' travel data, where the purpose of travel is to go to work. For each work activity, the start time is the time of arrival at the workplace, and the end time is the departure time of the next trip. If the trip is the last trip of the day, the end time of the corresponding work activity is recorded as 23:59 of that day.

[0019] Other activities are: all activities other than work activities and residential activities.

[0020] S12, Generation of rhythm curves for daily group activities.

[0021] S121. In terms of spatial scale, the collection area is divided into... A rule network; on a time scale, the entire day is divided into... There are 10 time intervals, and each time interval is 10 minutes.

[0022] S122. For any activity type, based on resident travel data, obtain the activity location of each individual under the corresponding activity type, and aggregate the activity locations in time and space. The group activity intensity is calculated by counting the number of people participating in group activities within each grid g and time interval k. A group activity intensity sequence is constructed using days as the time unit: ,in, Indicates the time interval Inside, in the grid Participate in the activity type The intensity of group activities; This indicates the total number of time intervals.

[0023] Specifically, let's assume the set of activity types is... For any activity type In each grid The daily activity rhythms of each group were constructed. Hypothetical variables... Represents an individual In time interval Does it appear in the grid? And it is in the activity type Then the grid In terms of activity type Next The intensity of group activity over a time interval is: .

[0024] S123, Pair each sequence recorded in the group activity sequence After normalization, the rhythm curve of the group's daily activities is obtained as follows: , Indicates the time interval Inside, in the grid Participate in the activity type The group's daily activity points.

[0025] In detail, the normalization process takes the following form: , in, This represents the minimum intensity of group activity; This represents the maximum intensity of group activity.

[0026] S2. Using activity type as the label for the daily activity rhythm curve of the corresponding group, construct a training set, and use the training set to train the initial activity type classification model to obtain the activity type classification model.

[0027] In this step, a multilayer perceptron model is constructed, using activity rhythm curves as input features and outputting the corresponding activity type category. The model consists of an input layer, several hidden layers, and an output layer, and uses a non-linear activation function to characterize the complex mapping relationship between activity rhythm features and activity types. During the training phase, the daily activity rhythms of the group are collected. and their corresponding activity type tags The multilayer perceptron model is used as a training sample to optimize its parameters through supervised learning, enabling the model to accurately depict the typical daily activity rhythm patterns corresponding to different activity types.

[0028] The activity type classification model obtained after training includes one input layer, two fully connected hidden layers, and one output layer. During model inference, the input of the input layer is set to the individual's daily activity rhythm curve. The number of neurons in the hidden layers are 64 and 32, respectively. The output layer uses the Softmax activation function to convert the output of the last hidden layer into the probability distribution of activity type.

[0029] S3. Construct an individual's daily activity rhythm curve based on mobile communication data, and output the individual's activity type using an activity type classification model based on the individual's daily activity rhythm curve.

[0030] S31. Extract individual activity points and activity records based on mobile communication data, and identify individual activity locations based on individual activity points.

[0031] Specifically, based on mobile communication data, strong spatiotemporal correlations are extracted from dwell points as activity points to identify the activity locations of individuals. This process is common knowledge to those skilled in the art, and the specific process can be referred to the technical solution provided in Chinese patent application CN118349867A, which will not be elaborated here.

[0032] S32. Project an individual's multi-day activity records in different activity locations onto a standardized 24-hour timeline, and discretize them into multiple time slices at the time level.

[0033] S33. For each individual's activity location, count its frequency of occurrence in different time segments, calculate the individual's cumulative activity intensity across days based on the frequency of occurrence, and construct an individual's cumulative activity intensity sequence across days for each activity location. , Indicates an individual's activity location Upper The cumulative activity intensity across days in a given time period, This indicates the total number of time slices.

[0034] We can assume that the individual identified a total of [number] individuals during the observation period. The number of activity venues and the number of observation days were [number missing]. The whole day is divided into Each time slice. Define indicator variables. Indicates the individual in the first... Heaven, the First Did they appear at the event location within that time frame? Then individuals in the activity area Upper The cumulative activity intensity across days for each time slot is: .

[0035] S34. Normalize the individual's cumulative activity intensity across days for each activity location in the individual's cumulative activity intensity sequence, and construct the individual's daily activity rhythm curve. , Indicates an individual's activity location Upper The individual daily activity rhythm curve points of each time slice.

[0036] In detail, the normalized form is as follows: , in, This represents the minimum cumulative activity intensity across days; This represents the maximum cumulative activity intensity across days.

[0037] This invention provides an activity type recognition system based on daily activity rhythms, used to implement the above-mentioned method, specifically including: Data Acquisition and Processing Module: This module collects resident travel data and mobile communication data. Based on the resident travel data, it identifies activity types according to activity type determination conditions. For each activity type, it generates a group daily activity rhythm curve based on the resident travel data and constructs an individual daily activity rhythm curve based on the mobile communication data. Activity types include residential activities, work activities, and other activities.

[0038] Classification Model Training Module: This module uses activity type as the label of the daily activity rhythm curve of the corresponding group, constructs a training set, and uses the training set to train the initial activity type classification model to obtain the activity type classification model.

[0039] The individual activity type generation module uses an activity type classification model based on the individual's daily activity rhythm curve to output the individual's activity type.

[0040] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0041] Furthermore, the present invention provides an electronic device including a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in read-only memory (ROM) or loaded from a storage unit into random access memory (RAM). Various programs and data required for device operation can also be stored in the RAM. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0042] Multiple components in the device are connected to the I / O interface, including: input units such as keyboards and mice; output units such as various types of displays and speakers; storage units such as disks and optical discs; and communication units such as network interface cards (NICs), modems, and wireless transceivers. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0043] The processing unit executes the various methods and processes described above, such as methods S1 to S3. For example, in some embodiments, methods S1 to S3 may be implemented as computer software programs tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via ROM and / or a communication unit. When the computer program is loaded into RAM and executed by the CPU, one or more steps of methods S1 to S3 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S3 by any other suitable means (e.g., by means of firmware).

[0044] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0045] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0046] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0047] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for identifying activity types based on daily activity rhythms, characterized in that, The methods include: Based on resident travel data, activity types are identified according to activity type determination criteria, and for each activity type, a daily activity rhythm curve of the group is generated based on the resident travel data; the activity types include residential activities, work activities, and other activities. The activity type is used as the label of the daily activity rhythm curve of the corresponding group to construct a training set. The initial activity type classification model is trained using the training set to obtain the activity type classification model. An individual's daily activity rhythm curve is constructed based on mobile communication data. Based on the individual's daily activity rhythm curve, an activity type classification model is used to output the individual's activity type.

2. The activity type identification method based on daily activity rhythm according to claim 1, characterized in that, The activity type determination criteria include: Based on the aforementioned resident travel data, if the starting location of a resident's first trip on a given day is their place of residence, then the activity type from 00:00 to the starting time of the first trip on that day is determined to be a residential activity; if the destination of a resident's last trip on a given day is their place of residence, then the time from the arrival time of the last trip to 23:59 on that day is determined to be a residential activity. Work activities are determined based on the travel logs recorded in the residents' travel data, where the purpose of travel is to go to work. For each work activity, the start time is the time of arrival at the workplace, and the end time is the departure time of the next trip. If the trip is the last trip of the day, the end time of the corresponding work activity is recorded as 23:59 of that day. The other activities mentioned refer to all activities other than work activities and residential activities.

3. The activity type identification method based on daily activity rhythm according to claim 1, characterized in that, The method for generating the aforementioned group daily activity rhythm curve is as follows: On a spatial scale, the data collection area is divided into a regular network; on a temporal scale, the entire day is divided into multiple time intervals. For any activity type, the activity location of each individual under the corresponding activity type is obtained based on the aforementioned resident travel data. These activity locations are then aggregated in time and space. The number of people participating in group activities within each grid and time interval is counted to determine the group activity intensity. A group activity intensity sequence is constructed using days as the time unit: ,in, Indicates the time interval Inside, in the grid Participate in the activity type The intensity of group activities; Indicates the total number of time intervals; Each sequence pair recorded in the aforementioned group activity sequence After normalization, the rhythm curve of the group's daily activities is obtained as follows: , Indicates the time interval Inside, in the grid Participate in the activity type The group's daily activity points.

4. The activity type identification method based on daily activity rhythm according to claim 1, characterized in that, The method for constructing the individual daily activity rhythm curve is as follows: Based on the aforementioned mobile communication data, individual activity points and activity records are extracted, and individual activity locations are identified based on the aforementioned individual activity points. The individual's multi-day activity records in different activity locations are projected onto a standardized timeline and discretized into multiple time slices at the time level; For each individual activity location, the frequency of its occurrence in different time slices is counted. Based on the frequency of occurrence, the cumulative activity intensity across days is calculated, and a cumulative activity intensity sequence across days is constructed for each activity location. , Indicates an individual's activity location Upper The cumulative activity intensity across days in a given time period, Indicates the total number of time slices; The cumulative daily activity intensity of each activity location in the individual's cumulative daily activity intensity sequence is normalized to construct an individual's daily activity rhythm curve. , Indicates an individual's activity location Upper The individual daily activity rhythm curve points of each time slice.

5. The activity type identification method based on daily activity rhythm according to claim 1, characterized in that, The activity type classification model includes an input layer, multiple fully connected hidden layers, and an output layer; wherein, the input of the input layer is an individual's daily activity rhythm curve; the number of neurons in each hidden layer is different; and the output layer converts the output of the last hidden layer into a probability distribution of activity type through a Softmax activation function.

6. An activity type recognition system based on daily activity rhythms, characterized in that, include Data Acquisition and Processing Module: This module collects resident travel data and mobile communication data. Based on the resident travel data, it identifies activity types according to activity type determination conditions. For each activity type, it generates a group daily activity rhythm curve based on the resident travel data. It also constructs an individual daily activity rhythm curve based on the mobile communication data. The activity types include residential activities, work activities, and other activities. Classification model training module: This module uses activity type as the label of the daily activity rhythm curve of the corresponding group, constructs a training set, and uses the training set to train an initial activity type classification model to obtain the activity type classification model; The individual activity type generation module outputs the individual activity type based on the individual daily activity rhythm curve and the activity type classification model.

7. The activity type recognition system based on daily activity rhythms according to claim 6, characterized in that, The activity type determination conditions performed by the data acquisition and processing module include: Based on the aforementioned resident travel data, if the starting location of a resident's first trip on a given day is their place of residence, then the activity type from 00:00 to the starting time of the first trip on that day is determined to be a residential activity; if the destination of a resident's last trip on a given day is their place of residence, then the time from the arrival time of the last trip to 23:59 on that day is determined to be a residential activity. Work activities are determined based on the travel logs recorded in the residents' travel data, where the purpose of travel is to go to work. For each work activity, the start time is the time of arrival at the workplace, and the end time is the departure time of the next trip. If the trip is the last trip of the day, the end time of the corresponding work activity is recorded as 23:59 of that day. The other activities mentioned refer to all activities other than work activities and residential activities.

8. The activity type recognition system based on daily activity rhythms according to claim 6, characterized in that, The method for generating the aforementioned group daily activity rhythm curve is as follows: On a spatial scale, the data collection area is divided into a regular network; on a temporal scale, the entire day is divided into multiple time intervals. For any activity type, the activity location of each individual under the corresponding activity type is obtained based on the aforementioned resident travel data. These activity locations are then aggregated in time and space. The number of people participating in group activities within each grid and time interval is counted to determine the group activity intensity. A group activity intensity sequence is constructed using days as the time unit: ,in, Indicates the time interval Inside, in the grid Participate in the activity type The intensity of group activities; Indicates the total number of time intervals; Each sequence pair recorded in the aforementioned group activity sequence After normalization, the rhythm curve of the group's daily activities is obtained as follows: , Indicates the time interval Inside, in the grid Participate in the activity type The group's daily activity points.

9. The activity type recognition system based on daily activity rhythms according to claim 6, characterized in that, The method for constructing the individual daily activity rhythm curve is as follows: Based on the aforementioned mobile communication data, individual activity points and activity records are extracted, and individual activity locations are identified based on the aforementioned individual activity points. The individual's multi-day activity records in different activity locations are projected onto a standardized timeline and discretized into multiple time slices at the time level; For each individual activity location, the frequency of its occurrence in different time slices is counted. Based on the frequency of occurrence, the cumulative activity intensity across days is calculated, and a cumulative activity intensity sequence across days is constructed for each activity location. , Indicates an individual's activity location Upper The cumulative activity intensity across days in a given time period, Indicates the total number of time slices; The cumulative daily activity intensity of each activity location in the individual's cumulative daily activity intensity sequence is normalized to construct an individual's daily activity rhythm curve. , Indicates an individual's activity location Upper The individual daily activity rhythm curve points of each time slice.

10. The activity type recognition system based on daily activity rhythms according to claim 6, characterized in that, The activity type classification model generated by the classification model training module includes an input layer, multiple fully connected hidden layers, and an output layer; wherein, the input of the input layer is an individual's daily activity rhythm curve; the number of neurons in each hidden layer is different; the output layer converts the output of the last hidden layer into a probability distribution of activity type through a Softmax activation function.

Images