Green travel influence factor mining and willingness degree measuring method based on machine learning

Abstract

The invention discloses a green travel influence factor mining and willingness degree measuring method based on machine learning, and belongs to the technical field of green travel, the green travel influence factor mining and willingness degree measuring method based on machine learning comprises the following steps: collecting a plurality of effective data according to set words, a Word2vec algorithm is used for carrying out screening; performing green travel willingness mining on the effective data by using a BERT model, and marking the grade of the green travel willingness of each piece of effective data; performing green travel influence factor extraction on the effective data based on LDA and UMAP; and carrying out association analysis on the green travel willingness and the influence factors. According to the method, text data related to green travel are crawled, text mining technologies such as a Word2vec algorithm are used for processing unstructured data, the screened data are analyzed, related effective data are screened, and an influence factor system of green travel willingness of residents can be objectively, truly and comprehensively extracted to a certain degree.

Description

Technical Field

[0001] The present invention relates to the technical field of green travel, and in particular to a method for mining green travel influencing factors and measuring willingness based on machine learning. Background Art

[0002] Statistics from the China Carbon Accounting Database (CEADs) show that from 1995 to 2021, my country's carbon emissions from the transportation sector increased from 130.21 million tons to about 727.9 million tons, a six-fold increase. In recent years, residents' excessive reliance on private cars has led to a rapid increase in the number of private cars, which has put great pressure on the city's transportation system and normalized events such as traffic congestion. At the same time, a series of problems such as air pollution and climate change caused by automobile exhaust emissions have caused great harm to health and the environment. Therefore, it is particularly urgent to solve the problem of carbon reduction in the field of transportation.

[0003] Green travel refers to the active use of transportation modes that reduce carbon dioxide emissions by individuals, including taking more urban rail transit, public buses and trams, carpooling, environmentally friendly driving, bicycles and walking, etc., also known as low-carbon travel or environmentally friendly travel. As a travel mode that saves energy, improves energy efficiency, reduces pollution, is beneficial to health, and takes efficiency into consideration, green travel can significantly alleviate urban traffic congestion, improve overall travel efficiency, reduce carbon emissions, and improve air quality, achieving sustainable use of environmental resources and sustainable development of transportation. In recent years, various countries have been actively and widely advocating green travel, and promoting the development of green travel has gradually been regarded as an important transportation planning goal.

[0004] Green travel is affected by various factors, including internal psychological factors, external situational factors, and individual heterogeneity. The internal psychological factors inherent to individual residents include attitudes, values, subjective norms, and behavioral intentions. External situational factors also affect residents' green travel, such as infrastructure construction and policy guarantees, social relations, low-carbon atmosphere, street greening, urban greening index, and architectural environment. In addition, there is an interactive effect between different factors.

[0005] There are many studies on the influencing factors of green travel behavior, but most of them use subjective methods such as literature review and grounded theory to extract factors, and then use highly structured data obtained from questionnaires for mechanism analysis. This mechanism analysis also uses traditional econometric models, ignoring the complex and overlapping correlations and causal relationships between influencing factors in the big data era. With the rise of social media and the development of big data technology in the Internet era, some scholars have begun to use multi-source data including unstructured data and machine learning methods to analyze the influencing factors and influencing mechanisms of environmental protection behaviors such as green consumption. Since a large amount of user text data in social media can be regarded as effective information about green lifestyles voluntarily provided by residents.

[0006] Although some studies have applied social media data to explore behavioral intentions or identify their influencing factors, most studies have not directly extracted relevant information on the level of action intention from unstructured posts. Instead, they still use scoring data from online comments or use questionnaires and other methods to obtain structured data for subsequent verification and analysis. The screening and extraction of unstructured data is not comprehensive enough, and the way of judging influencing factors is too single, making it difficult to reflect the correlation between influencing factors.

[0007] It should be noted that the information disclosed in the background technology section of the invention is only intended to deepen the understanding of the general background technology of the invention, and should not be regarded as an admission or suggestion in any form that the information constitutes prior art already known to those skilled in the art. Summary of the invention

[0008] The purpose of the present invention is to provide a method for mining green travel influencing factors and measuring willingness based on machine learning, so as to solve the problem that the screening and extraction of unstructured data is not comprehensive enough.

[0009] In order to solve the above technical problems, the present invention provides a method for mining green travel influencing factors and measuring willingness based on machine learning, comprising the following steps:

[0010] Collect several valid data according to the set words and filter them using the Word2vec algorithm;

[0011] Using the BERT model to mine green travel intentions on the valid data, and marking the level of green travel intentions for each piece of the valid data;

[0012] Extracting green travel influencing factors from the valid data based on LDA and UMAP;

[0013] Conduct a correlation analysis on green travel intention and influencing factors.

[0014] Preferably, before using the Word2vec algorithm for screening, the valid data is also preprocessed.

[0015] Preferably, using the Word2vec algorithm for screening includes: constructing a Skip-gram model, training the Skip-gram model using the valid data, setting the window to 10, using the Skip-gram model to calculate 50 words similar to the central word "green travel", and screening the valid data based on the 50 words using preset screening rules.

[0016] Preferably, the preset screening rule includes: if the sum of the cosine similarities of the 50 words contained in the text of the current valid data is greater than or equal to 1, the current valid data is retained; otherwise, the current valid data is deleted.

[0017] Preferably, the construction of the BERT model includes:

[0018] Provide several training data related to green travel intention;

[0019] Divide the training data into 5 levels from low to high according to the degree of green travel willingness, and mark them;

[0020] Using part of the training data as a training set and the remaining training data as a validation set;

[0021] Perform training and verification on the BERT model using the training set and the verification set;

[0022] Fine-tune using the Adam optimizer on the outer layer of the BERT model.

[0023] Preferably, the factors affecting green travel extracted based on LDA and UMAP include:

[0024] Using LDA to perform clustering processing on the valid data, and at the same time, providing each valid data with a statistical probability of belonging to each influencing factor theme;

[0025] Use the perplexity curve to determine the number of topics, check the keyword list in each cluster and optimize it;

[0026] The clustering results are visualized using the UMAP model.

[0027] Preferably, the correlation analysis between green travel intention and influencing factors includes: providing the statistical probability of the valid data belonging to each influencing factor theme as an independent variable, and using the level of green travel intention of the valid data as a dependent variable for analysis using a multiple linear regression model.

[0028] Preferably, the correlation analysis between green travel intention and influencing factors includes:

[0029] Using the ARM algorithm to extract a combination of influencing factors with a required value for the green travel willingness level from the valid data;

[0030] Establishing a Boolean data set using the statistical probability that each of the valid data belongs to each influencing factor theme;

[0031] If the statistical probability that the valid data belongs to the current influencing factor theme is greater than or equal to a set value, it is considered that the valid data has a corresponding influencing factor theme, otherwise, it does not exist.

[0032] A device for mining green travel influencing factors and measuring willingness based on machine learning, comprising:

[0033] The data acquisition module is used to collect a number of valid data according to the set keywords and filter them using the Word2vec algorithm;

[0034] A data processing module, used to mine green travel intentions from the valid data using a BERT model, and to mark the level of green travel intentions of each piece of the valid data. The data processing module also extracts green travel influencing factors from the valid data based on LDA and UMAP;

[0035] Data analysis module, used to conduct correlation analysis on green travel intention and influencing factors

[0036] A readable storage medium stores a computer program, which executes the above method when executed by a processor.

[0037] In the method for mining influencing factors of green travel and measuring willingness based on machine learning provided by the present invention, text data related to green travel is crawled, text mining technologies such as the Word2vec algorithm are used to process unstructured data, and the filtered data is analyzed to filter relevant valid data, so as to objectively, truly and comprehensively extract the influencing factor system of residents' green travel willingness to a certain extent; first, the BERT model is used to mine the behavioral willingness of residents for green travel, and the BERT text classification method is used to mine the residents' GTBI (green travel willingness) from the valid data, avoiding the need to re-design data collection tools to measure related variables in previous studies, and ensuring the measurement consistency of dependent variables and independent variables to a certain extent; secondly, an LDA topic model is constructed to extract factors affecting residents' green travel, and UMAP is used to visualize different factors on a two-dimensional plane.

[0038] The device for mining green travel influencing factors and measuring willingness based on machine learning provided by the present invention and the method for mining green travel influencing factors and measuring willingness based on machine learning provided by the present invention belong to the same inventive concept. Therefore, the device for mining green travel influencing factors and measuring willingness based on machine learning provided by the present invention at least has all the advantages of the method for mining green travel influencing factors and measuring willingness based on machine learning provided by the present invention, which will not be repeated here.

[0039] The readable storage medium provided by the present invention and the method for mining green travel influencing factors and measuring willingness based on machine learning provided by the present invention belong to the same inventive concept. Therefore, the readable storage medium provided by the present invention has at least all the advantages of the method for mining green travel influencing factors and measuring willingness based on machine learning provided by the present invention, which will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Those skilled in the art will appreciate that the accompanying drawings are provided for a better understanding of the present invention and do not constitute any limitation on the scope of the present invention.

[0041] Figure 1 This is a flow chart of a method for mining green travel influencing factors and measuring willingness based on machine learning according to an embodiment of the present invention;

[0042] Figure 2 It is a UMP visualization schematic diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, advantages and features of the present invention clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the drawings are all in a very simplified form and are not drawn to scale, and are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention. In addition, the structure shown in the drawings is often a part of the actual structure. In particular, the emphasis of each drawing is different, and sometimes different scales are used.

[0044] As used in the present invention, the singular forms "a", "an" and "the" include plural objects, the term "or" is generally used in a sense including "and / or", the term "several" is generally used in a sense including "at least one", and the term "at least two" is generally used in a sense including "two or more". In addition, the terms "first", "second" and "third" are used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second", and "third" may explicitly or implicitly include one or at least two of the features. The term "proximal end" is usually the end close to the operator, and the term "distal end" is usually the end close to the patient. "One end" and "the other end" as well as "proximal end" and "distal end" usually refer to two corresponding parts, which include not only the endpoints. The terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral body; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal connection of the two elements or the interaction relationship between the two elements. In addition, as used in the present invention, an element is arranged on another element, which usually only means that there is a connection, coupling, matching or transmission relationship between the two elements, and the connection, coupling, matching or transmission between the two elements can be direct or indirect through an intermediate element, and it cannot be understood as indicating or implying the spatial position relationship between the two elements, that is, one element can be in any position such as inside, outside, above, below or on one side of another element, unless the content clearly indicates otherwise. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0045] The inventors found that most studies did not directly extract relevant information on the level of intention to act from unstructured posts, but still used scoring data from online comments or used questionnaires and other methods to obtain structured data for subsequent verification and analysis.

[0046] Based on this, the core idea of ​​the present invention is to crawl text data related to green travel, use text mining technologies such as Word2vec algorithm to process unstructured data, analyze the filtered data, filter relevant valid data, and to a certain extent be able to objectively, truly and comprehensively extract the influencing factor system of residents' green travel willingness; first use the BERT model to mine residents' behavioral willingness for green travel, and use BERT's text classification method to mine residents' GTBI (green travel willingness) from valid data, avoiding the need to design data collection tools again to measure related variables in previous studies, and to a certain extent ensuring the measurement consistency of dependent variables and independent variables; secondly, construct an LDA topic model to extract factors affecting residents' green travel, and use UMAP to visualize different factors on a two-dimensional plane.

[0047] In addition, multiple regression was performed on the influencing factors and behavioral intention of green travel behavior to verify the effectiveness of the influencing factors, and an association rule algorithm was used to identify the combination of influencing factors for the group with high green travel behavior intention. By combining association rules and econometric models, the potential correlation and complex interaction mechanisms between various influencing factors were fully explored, and the focus of guiding and regulating green travel behavior was clarified.

[0048] For details, please refer to Figure 1-Figure 2 , which is a schematic diagram of an embodiment of the present invention. Figure 1 As shown, a method for mining green travel influencing factors and measuring willingness based on machine learning includes the following steps:

[0049] S1, collect several valid data according to the set words, and use the Word2vec algorithm to filter them.

[0050] In one embodiment, data is crawled using "green travel", "low-carbon travel", "environmentally friendly travel" and other green travel keywords, and specific green travel methods such as "bus", "bicycle", and "shared bicycle" as keywords. The crawled data of "green travel", "low-carbon travel", and "environmentally friendly travel" are mainly used, and the valid data crawled by integrating the keywords of specific green travel methods is used as the data set of the present disclosure.

[0051] Exemplary, for example, 261,116 Weibo posts from 2020 to 2022 are collected as valid data. It is understandable that since the crawled Weibo texts are highly non-standardized and unstructured, the valid data are also preprocessed before screening using the Word2vec algorithm.

[0052] Among them, the use of the Word2vec algorithm for screening includes: constructing a Skip-gram model, training the Skip-gram model using the valid data, setting the window to 10, using the Skip-gram model to calculate 50 words similar to the central word "green travel", and filtering the valid data based on the 50 words using preset screening rules. The Skip-gram model is used to train the results after the above preliminary preprocessing, with the window set to 10, and the 50 words in the corpus that are most similar to "green travel" are calculated. Here, only 50 words are taken as an example. Obviously, the number of words similar to "green travel" can also be adjusted according to the actual screening situation, and no specific limitation is made here.

[0053] Specifically, the preset screening rule includes: if the sum of the cosine similarities of the 50 words contained in the text of the current valid data is greater than or equal to 1, the current valid data is retained; otherwise, the current valid data is deleted.

[0054] For example, if the sum of the cosine similarities of the 50 words contained in a microblog text is greater than or equal to 1, it is considered to be a text related to this study; if it is less than 1, it is irrelevant. Irrelevant texts were deleted, and a total of 66,102 data sets were retained.

[0055] S2: Use the BERT model to mine the green travel intention of the valid data, and mark the level of green travel intention of each piece of the valid data.

[0056] Specifically, the construction of the BERT model includes:

[0057] Provide several training data related to green travel intention;

[0058] Divide the training data into 5 levels from low to high according to the degree of green travel willingness, and mark them;

[0059] Using part of the training data as a training set and the remaining training data as a validation set;

[0060] Perform training and verification on the BERT model using the training set and the verification set;

[0061] Fine-tune using the Adam optimizer on the outer layer of the BERT model.

[0062] In one embodiment, the training data here can be selected from the valid data or microblog texts selected by S1 above, or can be provided additionally. Here, only the valid data selected by S1 above is taken as an example. BERT mines the green travel behavior intention for each microblog text, and can classify the degree of green travel intention into 5 levels from low to high, 1-5, and manually annotates 5 types of training data, totaling 1,000 pieces, to be used as pre-training data to provide a standard for model willingness classification.

[0063] Part of the training data is used as a training set, and the rest of the training data is used as a validation set. Here, only 700 pieces are used as a training set and 300 pieces are used as a validation set. The Adam optimizer in the outermost layer of the model is fine-tuned to minimize the loss, and different hyperparameter combination experiments are conducted to obtain the best hyperparameter combination, thereby obtaining a good prediction output. For example, when epoch = 100, the accuracy of text classification reaches 82.6%. At this time, the validation loss rate is very close to the training loss rate, which indicates that there is no overfitting. Then, the classification results can be manually extracted and checked to ensure that the classification results output by BERT are consistent with the pre-defined classification standards, and can be used as a subsequent measurement model. It is determined that this is the optimal BERT combination model for this data set, which can be used to classify the remaining data for green travel behavior intentions.

[0064] In one embodiment, based on BERT classification, the recognition results of green travel behavior intention are shown in Table 1 below:

[0065] Table 1 GTBI classification results based on BERT

[0066] GTBI 1 2 3 4 5 Total Number 3949 6941 18458 16659 20095 66102 Proportion 5.97％ 10.50％ 27.92％ 25.20％ 30.40％ 1

[0067] Level 5 represents the highest GTBI, and its proportion of posts is the highest, reaching more than 30%. It is followed by general willingness (level 3) and higher willingness (level 4). Level 1, the lowest GTBI, has the least number of people, accounting for about 6% of the total posts. It can be seen that the user's GTBI level is relatively positive overall, which is more in line with the real world. On the one hand, this is because Weibo users are mostly young people aged 18-35. This group has a certain level of knowledge and has a certain degree of understanding of environmental protection concepts such as green travel and relevant national policies. On the other hand, the "Citizen Ecological and Environmental Behavior Survey Report (2022)" released by the Environmental and Economic Policy Research Center of the Ministry of Ecology and Environment on June 29, 2023 pointed out that the public's low-carbon travel situation is generally good; according to the "2023 China Major City Traffic Analysis Report" of Gaode Map, most urban residents have a high "green travel willingness".

[0068] S3, extracting green travel influencing factors from the valid data based on LDA and UMAP. Extracting green travel influencing factors based on LDA and UMAP includes:

[0069] Using LDA to perform clustering processing on the valid data, and at the same time, providing each valid data with a statistical probability of belonging to each influencing factor theme;

[0070] Use the perplexity curve to determine the number of topics, check the keyword list in each cluster and optimize it;

[0071] The clustering results are visualized using the UMAP model.

[0072] Among them, LDA is used to cluster the valid data, and the perplexity curve is used to determine the number of topics K in order to be more objective and accurate. After the clustering is completed, the keyword list in each cluster can be manually checked to check whether the clustering results are reliable and further optimize the clustering to obtain potential topics with similarities within the topics and characteristics that are obviously different from other topics.

[0073] In one embodiment, based on the LDA clustering algorithm, the optimal number of topics K can be set to 15 according to the perplexity curve. After checking the keywords, the remaining 14 clustering results can be combined into 11 different topics, i.e., 11 factors affecting residents' green travel behavior willingness, in combination with relevant opinions. According to the above processing, 11 factors affecting GTBI can be obtained, including psychological bias factors and situational factors, among which psychological bias factors are: attitude, perceived behavioral control, subjective norms-social media peers, subjective norms-social media stars and green knowledge; situational factors are: policy top-level design, economic policy, propaganda policy, restrictive policy, new energy technology and infrastructure.

[0074] In order to observe the influencing factor extraction results more clearly, the LDA topic results are further visualized. Since the feature vector of each text post has 512 dimensions, it is impossible to visualize it directly in 2D space. Therefore, the present disclosure uses the uniform manifold approximation and projection (UMAP) model to achieve visualization. It can be understood that after the effective data is reduced in dimension based on the LDA model, a low-dimensional data scatter plot is drawn using UMAP to visualize the above 11 categories of influencing factors in 2D space. As shown in the figure, different colors represent different topics, each topic has a clear boundary, and a single topic is clustered in a specific range, indicating that the clustering results based on LDA can distinguish topics well. It should be pointed out in particular that policy top-level design, subjective norms-social media companions, and subjective norms-social media stars are rarely involved in previous studies. This is also one of the theoretical contributions of the present disclosure. Figure 2 shown.

[0075] S4, conduct correlation analysis between green travel intention and influencing factors.

[0076] In one embodiment, the correlation analysis between green travel willingness and influencing factors includes: providing the statistical probability of the valid data belonging to each influencing factor theme as an independent variable, and using the level of green travel willingness of the valid data as a dependent variable for analysis using a multivariate linear regression model.

[0077] In S2 and S3, each microblog text has obtained the statistical probability of belonging to each influencing factor topic. After BERT classification, each microblog text is assigned a corresponding GTBI level (1-5) label. In order to test the relationship between the influencing factors extracted by LDA topic modeling and the GTBI level, in this part, the probability of each microblog text belonging to each topic is used as an independent variable, and the GTBI level of each microblog text mined by BERT is used as the dependent variable for multiple linear regression analysis.

[0078] In one embodiment, in order to fully explore the potential correlation between various influencing factors of the high GTBI population and clarify the focus of green travel behavior guidance and regulation, the correlation analysis between green travel intention and influencing factors includes:

[0079] First, the ARM algorithm is used to extract the influencing factor combination with the required value of the green travel intention level from the valid data, for example, the ARM algorithm is used to extract the influencing factor combination of the microblog text with a GTBI level of 5 from the processed data.

[0080] Exemplarily, 173,048 association rules were generated using ARM. Strong rules are selected by setting a threshold, which is determined based on the number of rules and the distribution of the three indicators. Rules with a lift greater than 2 and a confidence greater than 0.6 are defined as strong association rules, and a total of 10 strong association rules are generated, as shown in Table 2 below. Due to the limited computational complexity and computational resources of ARM, the association rules generated in this disclosure have a maximum of three factors on one side.

[0081] The results show that in the data set consisting of 20,095 microblog texts with a GTBI level of 5, the factors that lead to high GTBI are mainly perceived behavioral control, policy top-level design, green knowledge, restrictive policies and infrastructure. The most common combination of influencing factors is policy top-level design, green knowledge and infrastructure, with about 2,940 posts containing this factor combination, accounting for about 15%. In the above multivariate regression analysis, it is proved that when restrictive policy factors appear alone, they often lead to a decrease in GTBI, but the results of association rule mining show that when they appear together with other influencing factors, they will lead to a higher GTBI. The mutual embedding of influencing factors enhances their influence, which can better increase residents' GTBI compared with a single influencing factor, not only amplifying the effect of positive factors, but also mitigating the negative impact of restrictive policy factors. The conclusion of this study verifies the conclusion of (Stern, 1999), that is, it is not enough to use only a single influencing factor, and the combination of various factors is more effective, which has important implications for policy formulation.

[0082] Table 2 Strong association rules

[0083]

[0084]

[0085] Secondly, a Boolean data set is established using the statistical probability that each valid data belongs to each influencing factor theme. A microblog text can be regarded as a transaction, and the items in each transaction are a combination of GTBI and all influencing factors. It is necessary to establish a Boolean data set based on the valid data theme influencing factor-probability value obtained by extracting influencing factors from the previous part of LDA.

[0086] Finally, if the statistical probability of the valid data belonging to the current influencing factor theme is greater than or equal to the set value, it is considered that the valid data has a corresponding influencing factor theme, otherwise, it does not exist. For example, if the factor of the valid data theme influencing factor-probability value is greater than or equal to 0.04, it is considered to be 1, that is, this post has the theme; on the contrary, if it is less than 0.03, it is recorded as 0, that is, the factor does not exist, and the three indicators of support, confidence and lift are used as the standard for judging whether the association rule is a strong association rule.

[0087] Based on the same technical concept, the present disclosure also provides a device for mining green travel influencing factors and measuring willingness based on machine learning, including:

[0088] The data acquisition module is used to collect a number of valid data according to the set keywords and filter them using the Word2vec algorithm;

[0089] A data processing module, used to mine green travel intentions from the valid data using a BERT model, and to mark the level of green travel intentions of each piece of the valid data. The data processing module also extracts green travel influencing factors from the valid data based on LDA and UMAP;

[0090] The data analysis module is used to conduct correlation analysis on green travel intention and influencing factors.

[0091] Based on the same technical concept, the present disclosure also provides a readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method described above is executed.

[0092] The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0093] In the present invention, based on social media big data, a deep learning method is used to identify the degree of GTBI (green travel intention) from user-generated content. At the same time, an objective topic modeling method is used to extract the influencing factor system of residents' GTBI, and an association rule algorithm is used to explore the interesting correlation between the influencing factors and residents' GTBI. The value of green travel behavior big data is mined and previous literature data is integrated to form a multi-source data set. A comprehensive theoretical model of the influencing factors of green travel behavior intention is constructed, and the effective driving mechanism of urban residents' green travel behavior intention is explored.

[0094] The above description is only a description of the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any changes or modifications made by a person skilled in the art in the field of the present invention based on the above disclosure shall fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for mining green travel influencing factors and measuring willingness based on machine learning, characterized in that: The following steps are involved: Collect several valid data according to the set words and filter them using the Word2vec algorithm; Using the BERT model to mine green travel intentions on the valid data, and marking the level of green travel intentions for each piece of the valid data; Extracting green travel influencing factors from the valid data based on LDA and UMAP; Conduct a correlation analysis on green travel intention and influencing factors.

2. The method for mining green travel influencing factors and measuring willingness based on machine learning according to claim 1 is characterized in that: Before using the Word2vec algorithm for screening, the valid data is also preprocessed.

3. The method for mining green travel influencing factors and measuring willingness based on machine learning according to claim 1 is characterized in that: Using the Word2vec algorithm for screening includes: building a Skip-gram model, training the Skip-gram model using the valid data, setting the window to 10, using the Skip-gram model to calculate 50 words similar to the central word "green travel", and screening the valid data based on the 50 words using preset screening rules.

4. The method for mining green travel influencing factors and measuring willingness based on machine learning according to claim 3 is characterized in that: The preset screening rule includes: if the sum of the cosine similarities of the 50 words contained in the text of the current valid data is greater than or equal to 1, the current valid data is retained; otherwise, the current valid data is deleted.

5. The method for mining green travel influencing factors and measuring willingness based on machine learning according to claim 1 is characterized in that: The construction of the BERT model includes: Provide several training data related to green travel intention; Divide the training data into 5 levels from low to high according to the degree of green travel willingness, and mark them; Using part of the training data as a training set and the remaining training data as a validation set; Perform training and verification on the BERT model using the training set and the verification set; Fine-tune using the Adam optimizer on the outer layer of the BERT model.

6. The method for mining green travel influencing factors and measuring willingness based on machine learning according to claim 1 is characterized in that: The factors affecting green travel extracted based on LDA and UMAP include: Using LDA to perform clustering processing on the valid data, and at the same time, providing each valid data with a statistical probability of belonging to each influencing factor theme; Use the perplexity curve to determine the number of topics, check the keyword list in each cluster and optimize it; The clustering results are visualized using the UMAP model.

7. The method for mining green travel influencing factors and measuring willingness based on machine learning according to claim 6 is characterized in that: The correlation analysis between green travel intention and influencing factors includes: using the statistical probability of the valid data belonging to each influencing factor theme as an independent variable, and using the level of green travel intention of the valid data as a dependent variable to analyze using a multiple linear regression model.

8. The method for mining green travel influencing factors and measuring willingness based on machine learning according to claim 6 is characterized in that: The correlation analysis between green travel intention and influencing factors includes: Using the ARM algorithm to extract a combination of influencing factors with a required value for the green travel willingness level from the valid data; Establishing a Boolean data set using the statistical probability that each of the valid data belongs to each influencing factor theme; If the statistical probability that the valid data belongs to the current influencing factor theme is greater than or equal to a set value, it is considered that the valid data has a corresponding influencing factor theme, otherwise, it does not exist.

9. A device for mining green travel influencing factors and measuring willingness based on machine learning, characterized in that: include: The data acquisition module is used to collect a number of valid data according to the set keywords and filter them using the Word2vec algorithm; A data processing module, used to mine green travel intentions from the valid data using a BERT model, and to mark the level of green travel intentions of each piece of the valid data. The data processing module also extracts green travel influencing factors from the valid data based on LDA and UMAP; The data analysis module is used to conduct correlation analysis on green travel intention and influencing factors.

10. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is executed.

Images