In response to the three major challenges in the field of safety supervision in industrial parks: weak prevention, slow process response, and low efficiency of post review, this study analyzes the regional risk panorama of Longgang City and uses the "3030" hidden danger closed-loop mechanism in Longgang City as the research carrier to improve the efficiency of safety supervision and the accuracy of risk assessment in industrial parks through innovative methods. This article uses principal component analysis (PCA), analytic hierarchy process (AHP), and expert scoring method to determine the weights of seven types of risk factors in Longgang City, and constructs a regional risk panoramic analysis matrix model to evaluate the degree of regional risk in Longgang City. Introduce DeepSeek for intelligent risk analysis and construct a three in one model of "intelligent prediction, graded intervention, and dynamic optimization". By using deep learning methods to analyze 950 multi-level safety hazard inspection data at the provincial, municipal, and county levels from 2023 to 2025, a risk feature map with Longgang regional characteristics is constructed. This study identified 7 major categories of high-frequency hazards and 25 subcategories of high-frequency hazards in Longgang City, and revealed 9 composite risk overlay scenario threats; Strengthen supervision over high-frequency hidden dangers, promote the reduction of regional risk thresholds, and provide replicable basis for the next step of building an integrated regulatory paradigm of "intelligent center+institutional innovation".

1.1 Background

As a pilot area for China's new urbanization reform, Longgang City in Zhejiang Province has shown significant dual-core development in administrative management system reform and digital economy since it was upgraded from a town to a city in 2019. Geographically, it is situated at the strategic intersection of the Yangtze River Delta and the Haixi Economic Zone, where the 183.99km² administrative region supports the economic activities of 471,600 permanent residents. In 2024, the city's GDP reached 44.37 billion yuan, forming a unique industrial ecosystem centered around the printing and packaging industry. Longgang is one of the three major printing and packaging industry bases in China, with a comprehensive printing and packaging industry chain that covers materials, equipment, packaging printing, design, and cultural creativity. The city offers a wide range of printing services, including flatbed offset printing, gravure printing, letterpress printing, and digital printing. It is home to several national-level golden business cards, such as China Printing City, China Gift City, China Printing Materials Trading Center, and China Calendar Distribution Center. In terms of administrative management innovation, the city has implemented the 'city-managed community' flat governance model and reduced 60% of its institutions, creating the first county-level 'large department system' governance model in the country. However, in the field of production safety supervision, it faces common challenges such as an inverted ratio of government and enterprise supervision (few people handling many tasks), single supervision methods, slow response times, and insufficient information integration capabilities. Traditional methods of manual inspection and supervision have revealed three major structural issues in safety oversight: First, the mismatch between the scale of enterprises and the capacity for regulatory supply, such as inspecting 140 enterprises per capita in industrial and mining sectors, which makes it difficult to achieve comprehensive coverage in the short term; Second, the technological gap in identifying potential hazards, where professional limitations hinder the accurate identification of risks, leading to technical monitoring blind spots; Third, the absence of a closed-loop governance mechanism, where frequent hidden dangers identified by provincial supervision reports often rebound after rectification, failing to meet the urgent needs of modern industrial park safety oversight.

Longgang urgently needs to develop a regulatory system that leverages digital technology, particularly by introducing AI to build an intelligent early warning system. This system will enable dynamic monitoring and efficient handling of potential hazards. By using deep learning algorithms, the system can accurately identify risk characteristics and construct a risk situation analysis model to pinpoint the core safety risks in enterprises (locations). For the 3030 hidden dangers identified at the provincial（Within 30 days, investigate and deal with 30 major hidden dangers in various industries, and complete the rectification loop within 30 days）, municipal, and county levels, the system can precisely identify high-frequency risks and optimize the regulatory process through the PDCA cycle to ensure timely hazard elimination and effective preventive measures. This approach forms a new safety supervision model centered on data-driven intelligence, encompassing 'intelligent risk early warning, thorough hazard clearance, and full-process closed-loop management.' The system uses AI to create a 'risk circuit breaker threshold matrix,' which accurately identifies and dynamically quantifies core risks in industries such as mining and fire safety, forming a risk map for each sector. Based on regulatory trace data, the system creates enterprise safety profiles to promote precise and targeted governance. The introduction of AI not only enhances the efficiency of hazard identification but also achieves multi-dimensional risk warnings through data integration, optimizing resource allocation and ensuring precise and efficient regulation. This supports Longgang City in building a modern safety supervision system, providing theoretical breakthroughs and practical references for the modernization of county governance.

1.2. Security status

Since 2021, Longgang has experienced 30 production safety incidents, resulting in 27 fatalities. By industry, 13 incidents occurred in the industrial and mining sector, leading to 10 deaths; 7 incidents in construction, resulting in 7 deaths; 5 incidents in road traffic, resulting in 5 deaths; 2 incidents involving maritime and fishing activities, resulting in 2 deaths; and 3 incidents in other categories, resulting in 3 deaths. As an industrial city, Longgang has a higher incidence of industrial and mining accidents, construction accidents, and road traffic accidents. The mortality rate per billion yuan of GDP has been relatively low, at 0.024 in 2022,0.01465 in 2023, and 0.01352 in 2024, maintaining a low level. The majority of accidents are classified as general incidents, with no incidents resulting in 2 or more deaths. The main causes of accidents are falls from heights and object strikes, with a few cases involving collapses, mechanical injuries, electric shocks, and fires. For more details on the accident situation since 2021, see Figure 1.

Figure 1: Statistics of Safety Accidents in Longgang City (2021-2024).

In terms of industry supervision and law enforcement, the EMD is responsible for the supervision of four major industries:Non-coal mines, Metallurgical industry and Trade, Hazardous Chemicals, and Fireworks. In terms of comprehensive supervision, comprehensive supervision and law enforcement has not yet been truly implemented, mainly focusing on guidance, coordination, supervision and inspection, notification, and interviews.

2.1 Industrial structural risks

By the end of 2024, the economic entities of Longgang city show the following characteristics:

1) Large market volume. There are 121,700 registered business entities, one business entity for every four people.

2) There are many industrial parks (including small and micro parks), with nearly 5,000 industrial enterprises.

3) There are many small and micro enterprises. The total number of enterprises is over 29,000. In terms of enterprise composition: there are 505 large-scale enterprises (a decrease of 24 from 2023), accounting for only 1.74% of the total; small and micro enterprises account for 98.26% (including individual businesses). For example, in the printing and packaging industry, there are 40,085 market entities, employing nearly 130,000 people, with an average of just 3.2 employees per enterprise.

4) Dispersed industrial structure. The city's industrial structure is "low, small and scattered", mainly labor-intensive products, most of the enterprises are traditional printing industry, recycled cotton textile industry, etc., and the proportion of science and technology enterprises is less than that

5) The printing division is fine. The industry is mainly traditional packaging printing, and the output value of printing and packaging accounts for more than 60% of the total output value of Longgang. The industrial chain is complete and the professional division is fine.

6) Risk and hidden dangers are dense. There are many labor-intensive enterprises, front-line workers have high repetitive work, facilities and equipment are updated slowly, mechanical operation accidents or fire accidents are easy to occur.

2.2 Human resource risk

Fierce competition among peers. The competition among printing enterprises is fierce, the profit margin of enterprises is reduced, and the human resource input of enterprises is insufficient.

High employee turnover. The company's salary and benefits are average, which makes it less attractive to young local workers aged 18-35, leading them to move into the service sector, start their own businesses, or relocate to cities like Hangzhou and Xiamen. Additionally, the workforce includes many migrant workers with a high proportion of those with a junior high school education or lower, resulting in high turnover. This situation reduces the company's willingness to invest in training and poses potential safety risks.

Imbalance of the number of people. There are 5,235 ABCDEF talents in the city, but there are less than 300 highly skilled talents (class E and above), and only 479 technical talents in R&D and industrial positions, accounting for only 9.15%. The shortage of high-end talents, high-end R&D personnel and technical workers is obvious, which hinders the upgrading of the printing industry.

2.3 Investment momentum risk

Investment structure data shows that in 2024, industrial investment accounted for 29.5% of the city's total investment, a 0.1

To conduct a thorough analysis of seven types of risk factors—industrial structure risk, human resources risk, investment momentum risk, park operation risk, departmental supervision risk, enterprise clustering risk, and high-risk enterprise risk—and their proportions in the overall risk matrix, a regional risk matrix model is established. This model quantitatively evaluates the weight and impact of each risk factor, establishing a comprehensive regional risk analysis matrix model that ensures the total weight of all risks is 1. The model combines subjective and objective weighting methods, determining the weight distribution through expert scoring and data analysis. Due to insufficient data, some city-wide data are replaced with data from the new city.

3.1 Regional risk panoramic analysis matrix model

Indicator screening. Core parameters are selected according to data availability and policy sensitivity

Weight calculation. The stratified weight is assigned by the coefficient of variation method

Risk calibration. The industry baseline value and the historical data of the jurisdiction are used as double benchmark verification

Risk coupling. Establish the standardized processing model R=Σ (ωi×Ri)

A regional risk panoramic analysis matrix model containing 7 types of risk factors was constructed. The weights were determined by principal component analysis (PCA), Analytic Hierarchy Process (AHP) and expert scoring, ensuring that the total weight was 1. The weight allocation matrix for the 7 types of risk factors is shown in Table 4, and the values of each item in the regional panoramic risk analysis matrix model are shown in Table 5.

3.2 Regional risk panoramic analysis matrix model equation

The total risk equation of the regional panoramic risk analysis matrix model is as follows:

Where: βi is the weight value of the key quantitative index in each risk factor (0<βi<1>

ωi is the weight value of each risk factor (0 <βi <1>

Table 4: Sampled Data Statistics on Occupied Land Areas of SMEs in City-Wide Industrial Parks.

Table 5: Regional panoramic risk analysis matrix model with various numerical values.

3.3. Model results

Calculated:

The regional risk value is calculated to be 0.63907. Using Deepseek analysis and assigning weights to each risk factor (see Table 4), the calculated value is 0.6040935, a difference of 4.05%. The differences are in the risk values for industrial structure, human resources, investment momentum, enterprise clustering, and high-risk enterprises. Overall, the weights assigned by Deepseek to each type of risk are relatively reasonable, with the regional risk value being approximately 0.604, placing it within the high-risk range. Comparing the results from my own calculations and those from Deepseek, I found that the risks of high-risk enterprises, departmental supervision, park operations, and industrial structure are particularly high, requiring enhanced attention to mitigate these risks.

Among the four major risks, adjusting the industrial structure is challenging and requires significant time, human resources, and financial investment. The optimal solution is to enhance the intensity and frequency of supervision, which can significantly reduce the risk probabilities associated with high-risk enterprises, departments, and park operations. To achieve this, it is essential to understand the classification of potential hazards, particularly those that occur  frequently (i.e., those that are frequently and repeatedly identified during law enforcement inspections). To thoroughly investigate the distribution of various types of hazards in industrial parks, especially the high-frequency hazards, and to ensure the fairness of the research, we selected the '3030' mechanism hazards identified and addressed during the provincial and municipal supervision of Longgang  as the research subjects. We included 950 major and general hazards (including key issues) discovered by the provincial, municipal, and county-level supervision from August 2023 to January 2025 in our study sample, as detailed in Table 6 and Figure 3.

Table 6: Regional panoramic risk analysis matrix model with various numerical values.

Figure 3: Number of general and major hidden dangers in various fields of Longgang City.

4.1. Overall situation of hidden dangers

Classified by the entities responsible for identifying and addressing potential hazards, there are 319 at the county level, accounting for 33.58%; 504 at the Wenzhou city level, accounting for 53.05%; and 127 at the provincial level, accounting for 13.37%. Among these, 495 are major hazards: 256 were identified at the county level (mainly 86 from industrial and mining, 45 from urban operations, 37 from construction, and 36 from special equipment), accounting for 51.7%; 143 were identified at the Wenzhou city level (mainly 104 from industrial and mining), accounting for 28.9%; and 96 were identified at the provincial level (mainly 83 from fire safety), accounting for 19.4%. The focus of provincial supervision is on fire safety, while the city-level focus is on industrial and mining, and the county-level focus is on industrial and mining, gas, construction, and special equipment.

Classified by industry, there are 485 items in the mining and industrial sector, 125 in fire safety, 84 in construction, 77 in urban operations, 59 in special equipment, 41 in cultural tourism, 27 in maritime and fishing activities, 27 in hazardous chemicals, 22 in road traffic, and 3 in new business models (1 from the Municipal Economic Development Bureau and 2 from the Municipal Resources and Planning Bureau). By hazard level, there are 495 major hazards, accounting for 52.1% of the total, mainly concentrated in mining and industry (197 major hazards and 2 key issues, accounting for 40.2%), urban operations (50 hazards, accounting for 10.1%), special equipment (48 hazards, accounting for 9.7%), and construction (39 hazards, accounting for 7.88%). There are 455 general hazards, accounting for 47.9% of the total, mainly concentrated in mining and industry (286 hazards, accounting for 62.86%), urban operations (45 hazards, accounting for 9.9%), fire safety (35 hazards, accounting for 7.7%), hazardous chemicals (20 hazards, accounting for 4.4%), and cultural tourism (19 hazards, accounting for 4.2%).

As can be seen from the above data, major hidden dangers and general hidden dangers found at the provincial, municipal and county levels account for 50

The following conclusions can be drawn based on the research:

1. There are 7 types of risk factors in Longgang, and the four types of risk are high, such as high-risk enterprises, department supervision, park operation and industrial structure.
2. The risk value of Longgang area is 0.604, which is in the high-risk range.
3. By introducing DeepSeek intelligent risk analysis and building a three-in-one model of "intelligent prediction, graded intervention and dynamic optimization", we can effectively identify and control high-frequency hidden dangers and significantly reduce the regional risk threshold. By strengthening the management of easily regulated hidden dangers, the regional risk can be reduced by more than 10%-30%.