Method for generating television portrait based on big data

Abstract

The invention provides a method for generating a television portrait based on big data, and relates to the technical field of data processing, and the method comprises the steps: constructing a label system, obtaining label categories, determining a feature set corresponding to each label according to the label categories, carrying out the modeling of the label categories, carrying out the training through the feature sets, and obtaining a generation model corresponding to the label categories. And processing the preprocessed television data of the corresponding data source category by using the generation model corresponding to the tag category, generating the corresponding tag, and constructing the television portrait, so that the problem that an enterprise cannot extract valuable information from the television data is solved, and the method is suitable for generating the television portrait.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method for generating a TV portrait based on big data. Background Art

[0002] With the popularization of smart TVs, almost every household has a smart TV. Users generate a large amount of data through operations on the TV every day. This accumulated data is a valuable asset that enterprises are eager to mine. However, how to extract valuable information from this complex and heterogeneous data to guide production decisions has become an important challenge for enterprises. Summary of the Invention

[0003] Technical problem to be solved by the present invention: The present invention provides a method for generating a TV portrait based on big data to solve the problem that enterprises cannot extract valuable information from TV data.

[0004] Technical solution adopted by the present invention to solve the above technical problem: A method for generating a TV portrait based on big data includes the following steps:

[0005] S1. Preprocess the TV data, and the preprocessing includes data cleaning, data standardization, and data source classification;

[0006] S2. Construct a tag system to obtain tag categories;

[0007] S3. Use correlation analysis and principal component analysis to determine the feature set corresponding to each tag;

[0008] S4. Model for each tag category respectively, and use the feature set for training to obtain the generation model corresponding to the tag category;

[0009] S5. Use the generation model corresponding to the tag category to process the TV data of the corresponding data source category after preprocessing, generate corresponding tags, and construct a TV portrait.

[0010] Further, the data cleaning includes removing duplicate data and abnormal data; the data standardization includes unifying the data format and unit, and the data source classification includes TV device information, power-on and power-off logs, voice logs, music logs, viewing logs, and application data.

[0011] Further, the tag system includes basic attribute tags, behavior habit tags, interest hobby tags, commercial attribute tags, and user value tags.

[0012] Further, the basic attribute tags include geographical location, the behavior habit tags include the number of power - ons and the power - on duration, the hobby tags include actors, film and television types, and application preferences, the commercial attribute tags include membership purchases and single - piece purchases, and the user value tags include activity and commercial value.

[0013] Further, the generation models corresponding to the tag categories include decision tree models, random forest models, and neural network models.

[0014] Advantages of the present invention: The present invention provides a method for generating a TV portrait based on big data. By constructing a tag system, obtaining tag categories, determining the feature set corresponding to each tag according to the tag categories, modeling for each tag category respectively, and training using the feature set, a generation model corresponding to the tag category is obtained. Using the generation model corresponding to the tag category, the TV data of the corresponding data source category after pre - processing is processed to generate corresponding tags, and a TV portrait is constructed, solving the problem that enterprises cannot extract valuable information from TV data. Description of the Drawings

[0015] Figure 1 is a schematic flowchart of a method for generating a TV portrait based on big data provided by the present invention. Detailed Embodiments

[0016] Aiming at the problem that enterprises cannot extract valuable information from TV data, the present invention provides a method for generating a TV portrait based on big data, as Figure 1 shown, including the following steps:

[0017] S1. Pre - process the TV data, and the pre - processing includes data cleaning, data standardization, and data source classification.

[0018] Specifically, the data cleaning includes removing duplicate data and abnormal data; the data standardization includes unifying the data format and unit, and the data source classification includes TV device information, power - on and off logs, voice logs, music logs, viewing logs, and application data.

[0019] S2. Construct a tag system and obtain tag categories.

[0020] Specifically, the tag system includes basic attribute tags, behavior habit tags, hobby tags, commercial attribute tags, and user value tags. The basic attribute tags are usually static data, such as the geographical location where the TV device is located. The behavior habit tags include the number of power - ons and the power - on duration. The hobby tags include actors, film and television types, and application preferences. The commercial attribute tags include membership purchases and single - piece purchases. The user value tags include activity and commercial value.

[0021] S3. Use the correlation analysis method and the principal component analysis method to determine the feature set corresponding to each tag.

[0022] Specifically, for the basic attribute tags, the feature set includes the TV device information as input, the IP address as output, and the geographical location where the TV device is located obtained from the IP address. For the behavior habit tags, the feature set includes the power-on and power-off logs as input, and the number of power-on times and the power-on duration as output. For the interest hobby tags, the feature set includes the viewing logs as input, the actors and the types of movies and TV shows as output, and the application data as input, and the application preferences as output. For the commercial attribute tags, the feature set includes the viewing logs as input, and the membership purchase and single-piece purchase as output. For the user value tags, the feature set includes the power-on and power-off logs, voice logs, music logs, viewing logs, and application data as input data, and the activity and commercial value as output. The commercial value is obtained according to the user consumption behavior, and the consumption behavior includes the membership purchase behavior and the single-piece purchase behavior.

[0023] S4. Model for each tag category respectively, and use the feature set for training to obtain the generation model corresponding to the tag category.

[0024] Specifically, the generation model corresponding to the tag category includes a decision tree model, a random forest model, and a neural network model. Select a suitable machine learning model according to the feature set corresponding to the tag.

[0025] S5. Use the generation model corresponding to the tag category to process the TV data of the corresponding data source category after preprocessing, generate the corresponding tags, and construct a TV portrait.

[0026] Specifically, construct the TV portrait of a certain TV, that is, the tag set of this TV, to provide valuable information for the enterprise.

Claims

1. A method for generating a TV portrait based on big data, characterized in that, It includes the following steps: S1. Preprocess the TV data, and the preprocessing includes data cleaning, data standardization, and data source classification; S2. Construct a label system to obtain label categories; S3. Use correlation analysis and principal component analysis to determine the feature set corresponding to each label; S4. Build models for each label category respectively, and use the feature set for training to obtain the generation model corresponding to the label category; S5. Use the generation model corresponding to the label category to process the TV data of the corresponding data source category after preprocessing, generate corresponding labels, and construct a TV portrait.

2. The method for generating a TV portrait based on big data according to claim 1, wherein The data cleaning includes removing duplicate data and abnormal data; the data standardization includes unifying the data format and unit, and the data source classification includes TV device information, power-on and power-off logs, voice logs, music logs, viewing logs, and application data.

3. The method for generating a TV portrait based on big data according to claim 1, wherein The label system includes basic attribute labels, behavior habit labels, interest hobby labels, commercial attribute labels, and user value labels.

4. The method for generating a TV portrait based on big data according to claim 3, wherein The basic attribute labels include geographical location, the behavior habit labels include the number of power-on times and the power-on duration, the interest hobby labels include actors, film and television types, and application preferences, the commercial attribute labels include membership purchases and single-piece purchases, and the user value labels include activity and commercial value.

5. The method for generating a TV portrait based on big data according to claim 1, wherein The generation model corresponding to the label category includes a decision tree model, a random forest model, and a neural network model.

Images