JP-7855712-B2 - Information processing device

Inventors

• 相場 邦宏

Assignees

• 株式会社ＮＴＴドコモ

Dates

Publication Date

20260508

Application Date

20230719

Priority Date

20221027

Claims (5)

1. A waveform data generation unit generates waveform data for each action by determining the time-series timing of events related to the user's behavioral characteristics from behavioral data including the type of action and execution time information related to the acquired user's actions, and connecting adjacent timings in the time series with a predetermined waveform. An image generation unit generates a two-dimensional image representing the user's behavioral characteristics by converting the generated waveform data into a power spectrum image, Equipped with, The waveform data generation unit generates waveform data for each of a plurality of interrelated types of actions, and combines the obtained plurality of waveform data in a predetermined way to generate a single waveform data relating to the plurality of actions. Information processing device.
2. An estimation unit estimates the indicator by inputting a two-dimensional image representing the user's behavioral characteristics, generated by the image generation unit, as an explanatory variable into a learning model that uses user behavioral characteristic information as an explanatory variable and a predetermined indicator as an objective variable. The information processing apparatus according to claim 1, further comprising:
3. The aforementioned actions relate to the use of the service in question. The estimation unit estimates indicators related to the user's use of the service by using two-dimensional images representing the user's behavioral characteristics related to the use of the service, generated by the image generation unit, as explanatory variables. The information processing apparatus according to claim 2 .
4. The predetermined waveform includes at least one of a sine wave, a cosine wave, and a triangular wave. The information processing apparatus according to claim 1.
5. The power spectral imaging includes at least one of a continuous Fourier transform and a wavelet transform. The information processing apparatus according to claim 1.

Description

This disclosure relates to an information processing device that generates two-dimensional images representing user behavioral characteristics, and the generated two-dimensional images can be used, for example, as explanatory variables in machine learning. When using machine learning to estimate information about user behavior characteristics (dependent variable) from user behavior data (explanatory variables), it is important to generate and use explanatory variables with sufficient information to accurately estimate the dependent variable. Typically, user behavior data is input into the learning model as the total number of targets (e.g., behavior initiation) within a certain period (statistical value). Furthermore, Patent Document 1 below discloses a technique for generating pie charts and bar graphs that represent information about user behavior over time in order to easily grasp the type of behavior, duration of behavior, and quality of behavior over a given day. Japanese Patent Publication No. 2019-128736 This is a functional block diagram of the information processing device in the first and second embodiments.This is a flowchart showing the process in the first embodiment.This is a diagram to explain how to acquire behavioral data.(a) is a diagram showing the generation of waveform data related to the timing of app launches within the service, and (b) is a diagram showing the generation of waveform data related to the timing of purchase actions within the service.This figure shows an image of the actual data of the generated waveform data.This figure shows the generation of a power spectral image of a waveform.This figure shows an example of using the generated power spectral image as an explanatory variable.This is a flowchart showing the process in the second embodiment.This is a diagram to explain waveform synthesis.This figure shows the generation of a power spectral image from a composite waveform.This figure shows an example of using a power spectral image generated from a composite waveform as an explanatory variable.This figure shows an example of the hardware configuration of an information processing device. The following describes various embodiments of this disclosure with reference to the drawings. In the following, as the first embodiment, a basic embodiment is described in which waveform data for each action is generated by connecting the time-series timings of events related to the behavioral characteristics from the user's behavioral data, and a two-dimensional image representing the user's behavioral characteristics is generated by power spectral imaging of the individual waveform data. In the second embodiment, an embodiment is described in which processing is further carried out to group related waveform data from the multiple waveform data generated. Although "user behavior" includes various actions, in the following first and second embodiments, in order to obtain indicators that are formed over the medium to long term, such as the satisfaction level of the target service (hereinafter abbreviated as "service") for the user, "user behavior" will be described using user actions related to the use of the service as an example. (First Embodiment) As shown in Figure 1, the information processing device 10 in the first embodiment includes a behavior data acquisition unit 11, a waveform data generation unit 12, an image generation unit 13, and an estimation unit 14 as functional blocks for realizing the functions related to this disclosure. The functions of each unit will be outlined below. Details of the functions will be described in the explanation of the processing shown in Figure 2. The behavior data acquisition unit 11 is a functional unit that acquires behavior data, including the type of behavior and execution time information, related to the behavior of various users from external devices (for example, mobile terminals 20 of various users (Figure 3)), and stores the acquired user-specific behavior data in the built-in behavior database 11A. The waveform data generation unit 12 is a functional unit that generates waveform data for each action by determining the time-series timing of events related to the user's behavioral characteristics from the above-mentioned action data and connecting adjacent timings in the time series with a predetermined waveform. Examples of events related to the user's behavioral characteristics include launching an application within the service (hereinafter abbreviated as "app") and making a purchase within the service. Examples of the "predetermined waveform" that connects adjacent timings in the time series include sine waves, cosine waves, and triangular waves. The waveform data generation unit 12 may also have a function to group related waveform data from among the multiple waveform data generated, but this function of grouping waveform data will be explained in the second embodiment. The image generation unit 13 is a functional unit that gener