A data-driven high-speed railway service quality optimisation model and its influencing factors for grouping research

In the service quality of high-speed railways, the railway department needs to provide potential services in addition to the prescribed services for travellers. The prescribed services refer to the basic services that must be met under the requirements of railway service standards, i.e., the basic facility services and personnel services that satisfy the travelling needs of passengers, such as ticketing services, waiting for services, train services, station entry and exit services, and so on. Potential services refer to additional services that enhance passengers’ sense of well-being during travel when there is no clear requirement under the railway service specifications.

From the perspective of passenger perception, the performance of high-speed railway service quality perceived by passengers is shown in Figure 1, i.e., technical service quality and functional service quality. Technical service quality is mainly reflected in speed, safety, punctuality, and economy, while functional service quality is mainly reflected in comfort, convenience, and civilisation.

Figure 1.

The quality of high-speed railway service

Rapidity is reflected in high-speed railway passenger trains that can run at high speeds, reducing travel time for passengers and achieving a smooth and convenient experience during the journey. Safety is a kind of guarantee for travelers’ personal and property safety, and protects their personal life and property safety from being infringed. Punctuality is reflected in the strict control of the departure and arrival times of passenger trains by the railway authorities. The economy is embodied in the fact that when passengers receive the services provided by high-speed railways, they can enjoy better quality services within the reasonable consumption range of their psychological expectations. Technical quality refers to the level of service quality that can be observed in concrete terms. Comfort reflects that passengers enjoy warm and thoughtful service while traveling and meet basic requirements for comfort. Convenience is reflected in the travelers’ demand for convenient and fast service in the railway service. Civilization is reflected in high-speed railway transport enterprises through the provision of services to passengers and the degree of satisfaction of their spiritual needs.

In terms of the evaluation system of public service quality, the SERVQUAL evaluation model is the most suitable model for measuring the public service field, which is evaluated from the perspective of “travelers”, and the perceived value of “travelers” is taken as the standard for measuring the quality of public service. This model is based on the evaluation from the perspective of “travelers” and takes the perceived value of “travelers” as the standard to measure the quality of public services, which fully reflects the value of “travelers” [29]. The determinants of travelers’ perceived service quality are shown in Figure 2. The core of the theory is the “passenger perceived service quality gap model”, that is, the gap between the service expected by “travelers” and the actual perceived service.

Figure 2.

Determinants of customer perceived service quality

Existing research has addressed the determinants of the SERVQUAL evaluation model, i.e., tangibles, reliability, responsiveness, assurance and empathy, to develop a service quality measurement scale containing 22 representative evaluation indicators, known as the SERVQUAL Evaluation Scale.

According to the actual operating conditions and enterprise characteristics of high-speed railways, the theoretical indicators and evaluation methods related to the SERVQUAL model are applied as the evaluation method of high-speed railway service quality, and its evaluation framework is shown in Figure 3. Firstly, starting from the evaluation dimensions and indicators, the overall framework of service quality in this paper is built by constructing 5 dimensions and 22 indicators. According to the framework structure, and then combined with the actual high-speed railway service quality, the questionnaire is designed, and each high-speed railway station is selected to distribute the survey to random travellers. After collecting the questionnaires, the data is compiled, validity and reliability are tested, and the final statistical results are obtained. In this paper, the actual perception, expectation, and service quality gap of the multi-surveyed travelers are analysed in depth in detail. Based on the problems, relevant improvement suggestions are targeted.

Figure 3.

Service quality evaluation framework of High-speed railway

This paper constructs an evaluation index system around the SERVQUAL evaluation model based on the contents of the book “Service Quality Specification for High-speed Railway Medium-sized and Above Stations”.

The Specification for High-speed Railway Medium-sized and Above Stations is divided into nine chapters, which cover the scope of application, terminology and definition, passenger safety, equipment and facilities, civilised service, passenger organisation, commerce, advertisement management, basic management, and personnel quality, and it is a unified document regulating the service quality of high-speed railway stations.

The service quality evaluation index system of high-speed railway is shown in Table 1, including two levels of first-level and second-level indexes and the first-level indexes are the five dimensions of the SERVQUAL model, which are tangibility, reliability, responsiveness, assurance and empathy. Based on the Norms, the contents of the nine parts of the Norms are screened and categorised according to the SERVQUAL evaluation model, and the first-level indicators are refined and analyzed to obtain the second-level indicators.

Physicality, reliability, responsiveness, assurance and empathy are the five elements used by the SERVQUAL evaluation model to assess service quality. Physicality focuses on infrastructure development, reliability focuses on the accurate positioning of services, responsiveness focuses on catering to travellers’ preferences, assurance focuses on punctuality and stability, and empathy focuses on satisfying psychological needs. Based on the SERVQUAL evaluation index system established in the previous paper, this paper chooses the SERVQUAL model to evaluate the service quality of high-speed railways starting from the gap between travellers’ expectations and perceptions [30]. The calculation process is as follows:

According to the service quality of high-speed railways, the SERVQUAL scale is designed, and basic data on travelers’ expectations and perceptions of service quality are collected and obtained through questionnaire surveys. The calculation formula is: 1SQ=∑inPi−Ei Where SQ is the perceived service quality, P_i is the traveller’s perception score for the i rd factor, and E_i is the traveller’s expectation score for the i th factor.

According to the application of SERVQUAL model, SQ obtained by the above formula is the total value of perceived service quality of a single traveller under the condition that the five attributes of service quality are equal. In reality, travelers have different tendencies towards the five service quality attributes. Therefore, it is important to determine and weigh each attribute equally to obtain a more reasonable result. Specifically, it is the average difference between a traveller’s expectation and perception on an attribute multiplied by the weight of the attribute to obtain a traveller’s perceived service quality. That is: 2SQ=∑i=1nWi1m∑j=1m(Pij−Eij) Where, i represents the five major attributes of service quality, j represents each feature of the constituent attributes, P_ij represents the traveller’s expectation of the j th feature of the i th attribute, E_ij represents the traveller’s perceived value of the j th feature of the i th attribute, W_i is the weight of the i th attribute, and m represents the number of constituent features of the i th attribute.

Finally, in order to increase the accuracy of the data, it was taken to conduct a survey of multiple people, and the SQ scores of all the travellers who participated in the survey were added together and divided by the number of travellers K to obtain the average score for that service quality. I.e: 3Servqual =1K∑k=1kSQk Where Servqual is the perceived service quality, SQ_k is the perceived service quality of a single traveller, and k is the number of travellers participating in the survey.

To construct an optimisation scheme based on the service quality of high-speed railway, five aspects should be considered: tangibility, reliability, responsiveness, assurance and empathy. Where s is the indicator score and W is the combined weight of the indicator. The room for improvement of the indicator is reflected in the gap between the indicator and the average score s̅ of other trains in the same spectrum; the smaller the gap and the higher the average score, the less room for improvement. The difficulty of improving each indicator varies according to the actual situation of passenger management, this paper proposes the formula for calculating the recommended degree of indicator optimisation f , which takes into account the importance of the indicator and the room for improvement: 4f=W×(s¯−s)s¯

The indicator with the largest value of f corresponds to the service content that should be improved the most, and conversely, the indicator with the smallest value of f corresponds to the service item with the highest service quality in the same spectrum train, which is worth summarising the experience and providing reference for the service quality of high-speed railway. For each high-speed railway station under investigation, W can be obtained by the inverse entropy weighting method, and then combined with the scoring results of the passenger evaluation questionnaire to obtain the optimisation recommendation degree of each indicator.

Qualitative Comparative Analysis (QCA) is an analytical method that combines qualitative and quantitative analyses with group theory at its core. It transcends the division between quantitative and qualitative analyses while arguing that conditional variables do not influence outcomes independently, but rather by combining with other conditional variables to form groupings that synergistically influence outcomes.QCA is primarily used to address small- to medium-sized sample cases with group multiplicity, causal-conditional concurrency, group equivalence, and causal asymmetry by conducting cross-case comparative analysis. aiming to find out how each condition and condition group affects the occurrence of the outcome.

Fuzzy set qualitative comparative analysis (fsQCA) introduces the concept of affiliation, which is calculated by applying a fuzzy set algorithm based on the distance between the condition variable and the bipolar state (1 or 0) [31]. The condition variable is quantitatively assigned to any number between “0” and “1”, which can be 0.8, 0.6, 0.4, or 0.95, 0.75, 0.05, and so on. This is because many concepts in the objective world are mostly fuzzy phenomena, often can not be simply divided into 1 or 0, but when in a fuzzy state between 0 and 1.

The generative muscle composition analysis process of fsQCA is as follows: 1)

Fuzzy set affiliation calibration

QCA analysis is an approach based on set theory, which determines set relationships between measured variables by calibrating them. This method usually converts the variables into a set of sets and assigns a degree of affiliation to them [32]. An arbitrary mapping on a closed interval from the domain X to [0,1] is: 5μA:X→[0,1] 6x→μA(x)

Determine a fuzzy set A on X , μ_A is called the affiliation function of A , and μ_A(x) is called the degree of affiliation of x to the fuzzy set A , denoted as: 7A={ (x,μA(x))/x∈X }

In this paper, the “mean value anchor point method” for fuzzy set calibration, firstly, the variable data are arranged in descending order, and the quartiles are selected as the fully affiliated and fully unaffiliated values, and the mean value of the upper and lower quartiles is the value of the intersection point, so as to determine the anchor point. Then the anchor point values are expressed as follows:

Fully unaffiliated is: 8Q1=QUARTILE(x1,xn,1)

The intersection is: 9Q2=QQURTILE(x1j,xnj,2)

Fully affiliated as: 10Q3=QUARTILE(x1j,xnj,3)

The fuzzy set calibration can be expressed as: 11xj′= Calibrate (xj,Q3,Q2,Q1)

This study will comprehensively analyze the influencing factors and grouping paths of high-speed railroad service quality from the grouping perspective, with a view to enriching the existing research on the optimization and improvement of high-speed railroad service quality. This study takes the analytical framework of the SERVQUAL evaluation model as the research basis and constructs the theoretical model of service quality influencing factors of high-speed railroads, as shown in Figure 4. The traditional mainstream statistical method is based on the assumption that the variables are independent of each other and studies the “marginal effect” of individual variables on the dependent variable, ignoring the possible “chemical reaction” between the independent variables. The service quality of high-speed railroads takes more account of passengers’ psychological perception, i.e., it is a complex psychological reaction and physiological performance, and the antecedent conditions must be intricate and complex, so it is necessary to take a holistic perspective to study.

Figure 4.

The model of the quality group state of high-speed railway service

Based on this, this study integrates the elements of the SERVQUAL evaluation model to construct a theoretical model of the impact of high-speed railway service quality grouping. This is how the 5 antecedent conditions of tangibility, reliability, responsiveness, assurance, and empathy are linked and matched, and through what paths they influence the service quality of high-speed railways.

The five antecedent variables in this study are tangibility (Mat), reliability (Rel), responsiveness (Res), assurance (Rea) and empathy (Tra), and the outcome variable is high-speed railway service quality (HSRSQ). The scale is mainly based on the SERVQUAL evaluation scale designed in the previous section for the acquisition of relevant research data. A five-point Likert scale was used for measurement, with 1 to 5 representing strongly disagree, disagree, average, agree and strongly agree, respectively, which was filled in by the subjects according to their real feelings during travelling on the high-speed railway in terms of the degree of agreement with the question items.

In this study, the questionnaire was created on Questionnaire Star’s website and distributed. Data were collected at GZ South Station through WeChat and other online social platforms, as well as offline paper forms. A total of 368 questionnaires were obtained in this questionnaire research activity, excluding 30 invalid questionnaires with obvious regularity of answers, short response time and nonresearch subjects, and finally, 338 valid questionnaires were retained, with a validity rate of 91.85% for the questionnaire.

In order to effectively analyse the passenger evaluation of high-speed rail service quality, this paper chooses GZ South Station as the research object and collects high-speed rail service quality evaluation data through questionnaires. For the weights of high-speed rail service quality evaluation factor indicators, the on-site survey is used to obtain the evaluation weights of passengers, and the weights of tangible, reliable, responsive, guaranteed and empathetic high-speed rail service quality are obtained as 30%, 30%, 20%, 10% and 10%, respectively, after collection and collation. After determining the weight of each indicator, based on the previous SERVQUAL model perception service evaluation steps, the score of each evaluation indicator within the high-speed railway service quality evaluation factors is obtained, as shown in Figure 5.

Figure 5.

Scores of evaluation indicators for service quality

Since the most significant feature of service quality is to make travellers gain a sense of satisfaction, the degree of satisfaction of travellers with the service quality of high-speed railways is the most intuitive manifestation of evaluating the level of service quality. Accordingly, the following conclusions can be obtained by converting the obtained passenger service quality score into the form of passenger satisfaction, i.e., by using the score of passengers’ perceived service quality divided by the percentage of the score of passengers’ expected service quality: 1)

From the survey results, it is easy to see that the mean value of the overall service quality score of GZ South Station is -0.46 points. For easy understanding, the service quality score of high-speed railway is now converted into passenger satisfaction, i.e., passenger satisfaction = (average passenger perception/average passenger expectation)*100%. Combining the expected value and perceived value of each index in the figure, the average perception and average expectation of passengers are 4.27 points and 4.73 points, respectively. The passenger satisfaction with the overall service quality of GZ South Station is 90.28%, indicating that passengers are satisfied with the high-speed railway service quality at GZ South Station.

To sum up, the evaluation results of high-speed railway service quality truly reflect the current situation of high-speed railway service quality in GZ South Station, which is consistent with certain deficiencies in the current high-speed railway service quality in GZ South Station. It can be seen that the results of the survey and analysis, as well as the SERVQUAL evaluation results, are relatively accurate and credible.

Based on the evaluation results of high-speed railway service quality, combined with the calculation method of high-speed railway service quality optimisation recommendation degree given in section 3.1.2, the corresponding recommendation degrees of 22 evaluation factors in the evaluation index system are solved by combining different evaluation index weights. Figure 6 shows the optimisation recommendation degree of high-speed rail service quality indicators.

Figure 6.

Optimization of High-speed railway service quality index

As can be seen from the figure, in the optimization recommendation degree of high-speed railway service quality in GZ South Station, four service quality elements only need to maintain the current state, namely, responding to passengers’ requests at any time (C4), always treating passengers courteously (D3), answering passengers’ questions and answering them with all the answers (D4), and keeping in mind the passengers’ best interests (F4) with the optimization recommendation degree of 0. These services have already reached a higher level, and with the limited resources, Maintaining the status quo is sufficient. In addition, with limited resources, 11 service quality elements need to be improved, and their optimisation ratings are all less than -0.01. The order of improvement for the 11 service quality elements is A1>A2>F2>A4>B5>C1>D2>B4>B2>F1>D1, with the top three ranked with optimisation ratings of -1.205, -1.065, -1.061, which include service equipment and tools (A1), and keeping passengers’ best interests in mind (F4). Include service equipment and tools (A1), service equipment and tools pleasing to the eye (A2) and service marketing time convenient for travellers (F2). Therefore, GZ South Railway Station needs to focus on the optimal application of service equipment and tools in the subsequent optimisation of high-speed railway service quality and formulate a more reasonable service marketing time to meet passengers’ needs. The above conclusions objectively reflect the shortcomings and defects of high-speed railway service quality at GZ South Station, and are a great reference for the subsequent optimization of high-speed railway services at GZ South Station.

In this paper, based on the data obtained from the questionnaire, the mean values corresponding to tangibility, reliability, responsiveness, assurance, empathy and high-speed railway service quality scores are taken as the initial assignments of these antecedent conditions and outcome variables, respectively. Calibration refers to the process of assigning a set of affiliation scores to a case, and the direct calibration method uses a logistic function to assign an affiliation value to an original variable value based on three qualitative anchors, namely, full affiliation, crossover point, and full non-affiliation. The most important thing in the direct calibration method is to determine the three qualitative calibration thresholds, i.e., the full affiliation threshold, the intersection threshold, and the complete non-affiliation threshold. In this paper, drawing on existing relevant studies, all antecedent conditions and outcome variables are calibrated using upper quartiles of 0.75, 0.5 & 0.25. With the help of the PERCENTILE function in EXCEL, the calibration anchor points of each variable can be calculated, and then in the fsQCA software, through the function calibrate(x,n1,n2,n3) to enter the three calibration anchor points can be calculated in the results of the subordinate scores between 0 and 1. The variable calibration anchor points are shown in Table 2. It should also be noted during calibration that when the calibrated affiliation score of a variable for a case is 0.25 (neither biased affiliation nor unaffiliation), the case will be omitted from subsequent analyses due to its inability to be attributed to 0 or 1. Therefore, in this paper, we refer to the existing practice in the relevant literature of adding the constant 0.01 to all conditions where 0.5 affiliation occurs to ensure that the case will not be removed from the fuzzy set analysis.

In fuzzy set analysis, the necessity of individual variables must be tested before a constructive analysis can be performed. If a factor is necessary to lead to a particular outcome, it is a core condition and is not included in the subsequent analysis of the combination of conditions. When consistency is greater than 0.85, the corresponding condition is necessary. In QCA, coverage is used to determine the explanatory power of a variable or combination of variables. The calibrated data were imported into the fsQCA software to obtain the coverage and consistency level of each antecedent variable as a necessary condition for high-speed railroad service quality. The results of the one-factor necessary condition analysis are shown in Table 3, where “~” denotes logical “non-”, which means the non-existent state or lower level state. As can be seen from the table, the consistency level of the high-speed railroad service quality responsiveness variable (0.827) is the highest, which is less than 0.85, and is not a necessary condition for high service quality of high-speed railroad. This result confirms that the combination of multiple factors produces higher high-speed railroad service quality, and no single factor has a significant impact. The reasons for generating high service quality of high-speed railroads are complex, and the formation process is the result of multiple factors.

In order to make the results of the sufficiency condition study applying the fsQCA method more convincing and to reduce its randomness and sensitivity, it is necessary to carry out a robustness test on the results of the combinations affecting the service quality of high-speed railways. The research involved in this paper is based on the foundation of set theory, so the common methods of set theory are chosen to perform the robustness test. The two common methods of adjusting consistency thresholds and calibration thresholds are used to cope with the instability in the parameter setting process, in which adjusting the consistency thresholds is done by keeping other conditions unchanged, increasing the minimum consistency criterion from 0.82 to 0.86, and adjusting the calibration thresholds is done by keeping other conditions unchanged, and increasing the values of the initial thresholds for each variable of complete affiliation, intersection and complete unaffiliation from 0.75, 0.5 and 0.25 to 0.85, 0.6 and 0.3. The results of the intermediate solution for the group configuration after adjusting the benchmarks are shown in Table 6. The new histogram configurations, after raising the consistency threshold in the table, have no significant changes compared to the original configurations, and all the adjusted configurations constitute a subset of all the benchmark configurations, and this consistency proves that the solution terms are robust. The new histogram configurations, after adjusting the calibration criteria, are fully consistent with the original histograms, again indicating that the results of this paper are robust.