Practice and Effectiveness Evaluation of Community Policing Informatization in Social Management

The function of storing large amounts of unstructured data is the most basic function for a community policing information system. The system needs to store data from multiple sources, including: surveillance video, census, criminal records and other data sources. In addition, in addition to providing the function of reimbursement of expenses, the system also needs to provide the function of leadership and financial approval according to the actual needs of the enterprise in the business. Relying only on a set of database management system and network storage equipment, the traditional community policing information system has certain limitations in storage capacity and analyzing and calculating capacity, and its scalability is also poor. Therefore, when faced with the rapid growth of data, the public security organs often appear data growth brought about by the problem of slowing down the query speed. More seriously, the problem of data loss or damage often occurs. In order to deal with the storage problem of big data, this paper adopts the combination of data warehouse and HDFS distributed file storage system in the Hadoop ecosystem to complete the data storage function.

First of all, in the data warehouse, the data is organized according to the division of the keywords of the data, and these keyword fields tend to be more stable, and there will not be any change for a long time, for example, we put the policing data firefighting data as well as the entry and exit management data in different data partitions. It is conducive to correlation analysis according to the data partition.

Secondly, the preprocessing of data mining, such as data cleaning and noise reduction, is completed in the data warehouse. In-database data mining is used to populate a downstream analytical data mart that can be used by data mining and statistical modeling professionals to visualize complex patterns. For example, police use these patterns to identify potential suspects so that they can be targeted for control in a limited way. The use of in-database analytics allows for the automation of data mining in highly concurrent and highly scalable database architectures.

Finally, the data warehouse is used as the core of data analytics, where the master data is rationally maintained in the data warehouse. When the data warehouse is at the center of data governance and data cleansing, it can help make sense of all the information.

Big data sets present the characteristics of massive, high-dimensional, and sparse, however, the associations of data items continue to increase with the increase of data, so association rule mining has become one of the research cores in the field of data mining. In the field of big data, the goal of data mining is generally divided into two kinds of predictive and descriptive tasks. Predictive tasks are mainly obtained from the data set from the previously unknown and unmeasurable rules, such as valuable patterns, correlations, structures, or abnormal behaviors; descriptive tasks are to make a description of the laws that exist in the data set by using methods such as sequential patterns or correlation analysis.

In this paper, in the process of mining association rules for police data, the data set D is described as a set consisting of n transactions with unique transaction identifiers T_ID, D={T1,T2,…,Tn}$D = \left\{ {{T_1},{T_2}, \ldots ,{T_n}} \right\}$; where any transaction T consists of a number of items I_i in the item set I, where the item set I is defined as I={I1,I2,…,Im}$I = \left\{ {{I_1},{I_2}, \ldots ,{I_m}} \right\}$, and then the transaction T is a subset T ⊆ I of the item set I, symbolically denoted as T={I1,I2,…,Ii;i∈m}$T = \left\{ {{I_1},{I_2}, \ldots ,{I_i};i \in m} \right\}$.

Definition (1) support: the proportion of transactions with items A and B to all transactions, i.e., the probability of items A and B occurring at the same time, and its mathematical representation can be described as: 1sup(A→B)=P(A∪B)${\sup_{\left( {A \to B} \right)}} = P\left( {A \cup B} \right)$

Definition (2) Confidence: the proportion of transactions with item A that also contain item B, i.e., the probability of item B occurring given that item A occurs. Its mathematical representation can be described as: 2conf(A→B)=P(B|A)$con{f_{\left( {A \to B} \right)}} = P\left( {B|A} \right)$

Definition (3) association rule: if sup(A⇒B)≥minn_sup${\sup_{\left( {A \Rightarrow B} \right)}} \geq \min n\_\sup$, conf(A⇒B)≥min_conf$con{f_{\left( {A \Rightarrow B} \right)}} \geq \min \_conf$ then A⇒B is an association rule.

Definition (4) Candidate item set: to obtain frequent k – item set F_k, a join operation is performed for F_k−1 to generate candidate k – item set, which is denoted as C_k. Candidate C_k contains all frequent k – item sets, and the members of which can be infrequent item sets.

The join step of association rule mining generates a large number of candidate itemsets that consume storage space and computational resources, and how to efficiently and accurately discover all the frequent patterns is the core problem of the mining process. A simple association rule mining model is shown in Fig. 2.

Figure 2.

Mining process of traditional association rule algorithm

Discovering association rules is one of the core topics in the field of data mining, these rules can be effectively used to discover unknown relationships that can provide a basis for prediction and decision making.

The source of the case information data of this system is derived from the case information management system, SQL database format, using the case entry data of a unit for the whole year of 2021-2023. There are 964 cases in the case registration card and 975 persons in the subject registration card.

Before mining the case information set, the work of preprocessing the existing dataset has to be done first. Because there are some attributes in the case information dataset that are irrelevant or redundant to the final mining result, which are redundant attributes. Although these attributes into the mining process will not affect the results of the analysis, but it will increase the complexity of the algorithm, aggravate the burden of the server, and reduce the performance of the mining system, so it is necessary to delete these attributes before mining. By continuously removing unnecessary attributes, the dataset for data mining is finally obtained.

The main problems of the original case information involved in this paper are: 1)

There are a number of unfilled data or obvious errors;

To address the above problems, this paper carries out manual and computer-assisted preprocessing of case information. 1)

Data processing of vacancies

In the usual work, the database of various types of tables often appear in a large number of vacant information, such as in the “amount of money involved” in this field in the case of property invasion, due to the property and goods involved, if it is the loss of goods, there are often empty values, which will bring trouble for the data statistics. In this paper, according to the nature of the case with reference to the previous average value to fill in for them to make up the value.

1)

Decision tree follows a top-down approach to construct a decision tree from the set of training tuples and their associated class labels. First predefine the termination condition “N”, input the data set and give an accumulator i, initial value is assigned 0. First pruning, generate the decision tree [i]$\left[ i \right]$, record the decision tree [i]$\left[ i \right]$. Evaluate the decision tree, whether it satisfies the termination condition or not. If the termination condition is not met, continue pruning; if the termination condition is met, end. This recursion continues until the termination conditions are met.

1)

The association rule finds the frequent itemsets by iterative loop. Set the minimum support min_sup port and i = l. Scan all the attributes to get the set of first itemsets. Compare the first itemset with min_sup port to find the frequent ones and generate a candidate set, i plus 1. Calculate the support of the candidate itemset, and the itemsets with greater than the minimum support are gathered together for the second loop. This is repeated until the frequent itemsets do not satisfy the support.

After the above data collection and conversion, the case information data is assembled in the database of the SQL ser ver 2005 server. The SQL server 2005 BI data mining tool will be able to access data from it.

Once the algorithm module is designed, the data mining model is built and the following step is to provide the data to be analyzed to the data mining engine. In this training phase, the system starts studying the input data with the data mining algorithm. The training set is scanned about once or a number of times in a cycle, from which correlations between them are found and memories are generated. The data mining tool applies the training mining model to the new dataset and predicts the predictable columns of each new instance according to the rules and calculates the possible values of these predictable columns.

This concludes the design process of the system mining algorithm module.

Before the beginning of the practice of community policing informatization, in order to understand the regularity and universality of the problems existing in the construction of urban community policing informatization in the city, this paper firstly adopts an anonymous online questionnaire to conduct a random survey on different types of community policing offices in the city. The residents in 15 communities were surveyed and the questionnaire survey was conducted for the effect of their community policing work, the total number of questionnaires issued, the number of recovered questionnaires and the effective rate of questionnaires were 500, 432, and 86.4%, respectively.

The questionnaire design content is roughly divided into two parts, the first part includes the basic information part and the community policing informatization construction satisfaction survey part, mainly through the survey to understand the residents’ satisfaction with the city’s urban community policing informatization construction, so as to intuitively understand the residents’ perception, a total of 8 multiple choice questions. The second part of the survey mainly focuses on the theme of community policing informatization construction, analyzing the participation of community residents in community policing, with a total of three questions.

There are five questions about the perception of community policing by community residents: “Do you know that there is a police office in the community” (Q1), “Do you know the responsibilities of the community police office” (Q2), “Do you know the telephone number of the community police office” (Q3), “Have you ever asked for help from community policing (Q4)”, “Are you satisfied with community policing services (Q5)”. Details of the survey results are shown in Figure 3. From the information of the relevant data in this figure, it can be seen that the percentage of those who do not know whether there is a community policing office, do not know the specific duties of the community policing office, and do not know the telephone number of the community policing are 38.25%, 71.93%, and 57.46%, respectively. In general, community residents have a low level of knowledge about community policing. The percentages of those who have not sought help from community policing and those who are not satisfied with community policing services are 61.53% and 78.34%, respectively, and the overall satisfaction level of community residents with community policing is low.

Figure 3.

Perceptions of community residents on community policing work

Regarding the participation of community residents in community policing, three questions were set up, including channels of participation in community policing (Q6), matters (Q7), and perceived problems of community policing (Q8), and the results of the survey are shown in Table 1, 51.39% of the community residents have not participated in community policing, and most of the others who have participated in it are because of publicity and training such as firefighting, which accounted for 61.11%, and 53.94% of the residents believe that community policing information collection is slow.

From the results of the questionnaire, it can be concluded that the residents of this community are not satisfied with the work of community policing, and the construction of information policing computerization in this community needs to be improved.

Mining and analyzing the origin and age structure of offenders whose cases are burglaries by Microsoft decision tree algorithm.

Select Accent in the tree drop-down list in the toolbar to get the “hometown” decision tree of the offender. Firstly, the algorithm branches the age range according to the calculation result of information entropy. When the Age Area=“30-40” node is selected, the proportion of the “hometown” of the offender between the ages of 30 and 40 is displayed. The decision tree of theft is shown in Figure 4, and it can be seen from Figure 4 that there are 264 cases belonging to the Age Area=“30-40” node, and the probability of “hometown=Hebei” is 51.52%. You can then continue to expand the tree until you find the node information of interest and finally form a rule.

Figure 4.

Theft decision tree

The Microsoft association rule algorithm model is used to explore the intrinsic connection between attribute items such as age range, place of origin, crime characteristics, and means of committing crimes, with a view to exploring the deep-seated crime laws and discovering the characteristics of the suspects, so as to assist in preventing and combating the crime.

First set the algorithm parameters of Microsoft association rule model, where Minimum_Support=30, Minimum Probability=0.5, Maximum Itemset Size=0. After the model is processed by SOL Server Analysis Services, the association rule viewer can be used to view the content of the model. Rule Viewer to view the contents of the model. The Association Rule Viewer contains three tabs: Rules, Itemsets, and Dependency Networks.

Based on the association rule model, 315 rules were generated and the results are shown in Table 2. It can be found that the criminals of the theft cases are mostly from Hebei and Jiangxi, and their ages are mostly between 30-50, and most of them are gangs of roving criminals. Therefore, the public security police should strengthen the inventory of the people who meet the above combination of conditions, and crack down on the crime of theft.