CN115878689A - Meteorological and urban operation sign association rule mining method based on grid management - Google Patents

Abstract

The invention discloses a method for mining correlation rules of weather and urban operation signs based on grid management. In addition, the invention also introduces a frequent pattern mining algorithm FP-Growth, further carries out mining of association rules of meteorological conditions and event occurrence, and constructs a typical event knowledge graph covering the meteorological conditions. By identifying the correlation characteristics of weather and event data, the deep fusion of urban operation big data and weather big data can be realized, and technical support and decision reference are provided for urban accurate management.

Description

Meteorological and urban operation sign association rule mining method based on grid management

Technical Field

The invention relates to the technical field of meteorological analysis, in particular to a method for mining correlation rules of meteorological signs and urban operation signs based on grid management.

Background

The fine treatment represents the future treatment direction of the city and is also an important challenge for constructing the prominent global city of the city. The current urban grid management patrols the matters in the responsibility grid by subdividing urban management units, setting special mechanisms, unifying working standards, assigning grid patrolmen and the like, and transmits the found problems to a disposal department through a specific information system for disposal, thereby becoming an important practical means for strengthening basic level construction, perfecting community management and realizing social management, socialization, legal management, intellectualization and specialization. The novel city management mode truly moves the center of gravity of the social management downwards to the basic community, and the refinement degree and the timely response degree of the management are greatly improved in the modes of management sinking, resource integration and block subdivision. Although the current grid management has the mechanism advantage of real-time monitoring and focuses on the presetting and processing of social security emergencies in urban management, the use efficiency and the effect of post-event management are still insufficient, and grid patrol data serving as important support data for urban operation and urban fine control are necessary to be fully utilized to further extract the data value.

At present, some scholars develop analysis research on weather and city operation related data, wherein Yang Chen and the like adopt a natural language processing method to extract 110 alarm disaster information and analyze disaster-causing weather conditions; ren Yongjian and the like develop sensitivity analysis of meteorological factors on summer maximum power loads, and establish a prediction model of Wuhan summer maximum power loads; bi and the like adopt a data driving method to carry out the influence analysis of weather on the urban traffic condition; park et al analyzed the relationship between air temperature and traffic accidents. In addition, many scholars also develop academic research aiming at the urban gridding management innovation mode and management mechanism and grid hotline data analysis based on a data science method. In the existing research, chang Yanjun and the like use an ARIMA model to carry out trend prediction on gridding management data; peng Xiao et al, based on "12345" citizen service hotline, develop analysis of time, space and type characteristics of citizen's incoming calls, and analyze time variation characteristics and space distribution patterns of various problems; wang Jieyi further realizes the mode mining and prediction of specific category events through the aggregation extraction of city management events.

The grid management data has a good indication significance for city operation signs, but the existing research has a great defect in revealing the time-space regularity of city management events on a fine scale, and weather conditions often affect the occurrence and development of the events, but the existing research is lack of the regular analysis of the weather conditions on the occurrence of the events.

Disclosure of Invention

In order to overcome the defects in the prior art, the invention provides a method for mining the association rule of weather and urban operation signs based on grid management big data, which can realize the deep fusion of the urban operation big data and the weather big data, identify the association characteristics of the weather and event data by developing the mining of the association rule of the weather conditions and grid events, construct a typical event knowledge graph covering the weather conditions, and provide technical support and decision reference for the accurate management of cities.

The invention is realized by the following technical scheme:

a method for mining association rules of weather and city operation signs based on grid management is characterized by comprising the following steps:

step 1, data selection and source

The data information comprises urban grid patrol data and synchronous meteorological observation data, the grid patrol data is obtained in real time through a data interface provided by a building department, the meteorological observation data is from an automatic meteorological observation station distributed in the whole city range by a meteorological bureau, and four meteorological elements of precipitation, air temperature, wind direction and wind speed are selected to participate in analysis;

step 2, data processing

Under the python environment, converting longitude and latitude into plane coordinates by adopting universal transverse-axis mercator projection under a geodetic coordinate system, extracting events in a buffer area in a certain range around the automatic meteorological station as an analysis target, and carrying out meteorological element assignment on the events;

step 3, data research

Performing time-space characteristic analysis and type characteristic analysis of events under R and python language environments based on grid patrol data and meteorological observation data, and constructing a keyword cloud and a co-occurrence term network by adopting a Chinese word segmentation and keyword extraction method through event description information;

by matching the time and the position of the event and the weather information, mining the association rule of the weather condition and the typical event by using an FP-Growth algorithm to obtain the corresponding matching rule, and carrying out graphical expression by using a knowledge graph.

The invention has the further improvement that the Chinese word segmentation method comprises the following steps: selecting a JeibaR word segmentation packet in an R language environment to perform word segmentation processing on a grid description text, and performing the word segmentation processing by adopting a mixed word segmentation engine in combination with a maximum probability method and a hidden Markov model.

The invention is further improved in that the keyword extraction method comprises the following steps: on the basis of segmenting the grid description information, a TF-IDF algorithm is adopted to extract keywords corresponding to the events.

The invention is further improved in that the step of mining the association rule of the meteorological condition and the typical event occurrence by using the FP-Growth algorithm comprises the following steps: selecting statistics of 6 hours, 12 hours and 24 hours of accumulated precipitation, the highest air temperature, the average air temperature, the lowest air temperature, the maximum wind speed and the average wind speed to construct meteorological features, dividing the meteorological features into different intervals, introducing non-meteorological features to construct features, mining frequent patterns between meteorological conditions and events through an FP-Growth algorithm, analyzing event types with high correlation degree with the meteorological conditions, large occurrence quantity and obvious increment, mining potential rules in the meteorological conditions and event data according to the frequent patterns, and quantifying the reliability of the rules.

In a further improvement of the invention, the step of graphically expressing by means of a knowledge-graph comprises: and mining results based on the FP-Growth frequent pattern, and processing the subset superset to complete a knowledge graph rule base and present a link relation between weather and events.

Due to the adoption of the technical scheme, the invention has the beneficial effects that:

the method comprises the steps of firstly, mining the characteristics of gridding event data by adopting a space-time analysis and natural language processing method, and on the basis, mainly combining meteorological observation data to carry out analysis and research on the influence of meteorological conditions on the occurrence of urban gridding management events. In addition, the invention also introduces a frequent pattern mining algorithm FP-Growth, further carries out mining of association rules of meteorological conditions and event occurrence, and constructs a typical event knowledge graph covering the meteorological conditions. By identifying the correlation characteristics of weather and event data, the deep fusion of urban operation big data and weather big data can be realized, and technical support and decision reference are provided for urban accurate management.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a yearly distribution plot of event numbers.

Fig. 2 is a daily distribution plot of events.

Fig. 3 is an event time-by-time profile.

FIG. 4 is a co-occurrence lexical network of events.

FIG. 5 is a knowledge graph of weather conditions and typical event occurrences, using a "traffic pole" as an example.

Fig. 6 is a knowledge graph of weather conditions and typical event occurrences, using "river pollution" as an example.

Detailed Description

The fine treatment is the future treatment direction of cities and is also an important challenge for constructing prominent global cities in cities. Most of the existing researches are based on urban grid management innovation modes and management mechanisms, but the analysis and mining of grid patrol event data have great defects and lack of the regular analysis of meteorological conditions on event occurrence. The urban operation management big data are analyzed and researched based on a space-time characteristic analysis method, a word cloud and co-occurrence word item characteristic analysis method, a correlation analysis method and a frequent pattern mining method, so that typical weather conditions triggering grid management events are obtained, and a typical event knowledge graph covering the weather conditions is constructed. The result shows that the event occurrence time is highly consistent with the working time, the occurrence area is also consistent with a dense region of urban personnel, the phenomenon of head concentration and long tail distribution exists in the category, a clear clustering structure can be formed on the event participle, and a co-occurrence lexical item relation network taking the citizen activities as the main body is formed. By combining meteorological data analysis, the relevance between the subclasses of municipal facilities, sanitation and the like and the air temperature is obvious, the wind-vulnerable structure is greatly influenced by wind power, and in the specific weather conditions of precipitation, low temperature, high temperature, strong wind and the like, events such as foundation pits, disputes, high-altitude parabolas, river greening and the like show a high-incidence trend. In addition, the method and the system induce and express the association between the weather and the urban operation by adopting the knowledge map technology, thereby being beneficial to the advance handling and disposal of urban operation managers under specific weather conditions.

The following further describes embodiments of the present invention with reference to the drawings. It should be noted that the description of the embodiments is provided to help understanding of the present invention, but the present invention is not limited thereto. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

The invention relates to a method for mining correlation rules of weather and city operation signs based on grid management, which mainly comprises the following steps:

step 1, data selection and source

The data information mainly comprises urban grid patrol data and contemporaneous meteorological observation data, the grid patrol data is obtained in real time through a data interface provided by a building department, and the data fields comprise event ID, discovery date and time, event description, event major category, minor category, subclass, affiliated area, street, address, longitude and latitude and other information. The meteorological observation data come from automatic meteorological observation stations distributed in the whole city range by the meteorological bureau, the elements comprise information such as time, station number, precipitation, air temperature, air pressure, relative humidity, wind direction, wind speed and visibility, and the time resolution is 1 hour. Because the correlation of the relative humidity with the air temperature and the rainfall is strong, and the number of visibility observation stations is limited, the research requirements cannot be met, four meteorological elements of the rainfall amount, the air temperature, the wind direction and the wind speed are mainly selected to participate in analysis.

Step 2, data processing

The method takes the automatic weather stations in the city as a reference, converts longitude and latitude into plane coordinates by adopting universal transverse axis mercator projection (UTM 51N) under a WGS84 geodetic coordinate system in a python environment, extracts events in a buffer area around the automatic weather stations, takes the events as an analysis target, and assigns meteorological elements to the analysis target, wherein the selected meteorological elements comprise 6-hour, 12-hour and 24-hour accumulated precipitation, the highest air temperature, the average air temperature, the lowest air temperature, the maximum wind speed and the average wind speed, and thus an analysis data set combining event types and weather conditions is obtained.

Step 3, data research

Because the grid patrol data has complete information description and position record, the invention performs spatiotemporal feature analysis and type feature analysis of events in R and python language environments based on the grid patrol and meteorological observation data, and constructs keyword cloud and co-occurrence lexical item network by adopting a Jieba Chinese word segmentation engine and a TF-IDF keyword extraction method through event description information. On the basis, the time and the position of the event and the weather information are matched, the FP-Growth algorithm is further utilized to mine association rules of weather conditions and typical events, corresponding matching rules are obtained, and graphical expression is carried out through a knowledge graph.

The Chinese word segmentation is a process of recombining Chinese linguistic data into a word sequence according to a certain rule. The invention selects a Jiebar word segmentation packet under an R language environment to perform word segmentation processing on a grid description text, and adopts a mixed word segmentation engine, namely a mode of combining a maximum probability method and a hidden Markov model. Because the effect of the word segmentation method has great relation with the word segmentation dictionary and the stop word processing, the invention adds the stop word aiming at the characteristics of the grid description information on the basis of the word segmentation dictionary carried by the Jiebar and carries out the word segmentation processing on the basis of the stop word dictionary.

On the basis of segmenting the grid description information, a word Frequency-Inverse Document Frequency (TF-IDF) algorithm is adopted to extract keywords corresponding to the event, and the TF-IDF can identify terms which are more frequently appeared in the grid description information and are rarely appeared in other text corpora, so that the keyword information corresponding to the event is highlighted. The inverse document frequency for any given term is defined as:

wherein n is _documents Is the total number of documents in the corpus, n _{documents containing term} Is the number of documents containing the term.

A Frequent Pattern Growth (FP-Growth) algorithm is an important application of a data mining technology in the field of association rule discovery, a database transaction with an item set meeting the minimum support degree is compressed into a Frequent Pattern Tree (FP-Tree) by adopting a divide-and-conquer strategy, association relation among the same transactions is reserved, a corresponding condition FP-Tree is found out according to each Frequent 1 item set in an item head table, and a Frequent item set is excavated until all the condition FP-Trees are excavated. Compared with other association rule mining algorithms, the FP-Growth algorithm has the remarkable characteristic of high efficiency, and frequent item sets in a data set and the association rules hidden in the data can be efficiently discovered. Because the algorithm aims to mine a mode frequently appearing in a specific meteorological condition and a specific event, interference caused by randomly-occurring non-meteorological factors can be eliminated to a certain extent. Therefore, when the weather condition and event correlation rules are mined, the algorithm is adopted to analyze the weather and the events.

In the research, statistical values of 6 hours, 12 hours and 24 hours of accumulated precipitation, the highest air temperature, the average air temperature, the lowest air temperature, the maximum wind speed, the average wind speed and the like are selected to construct meteorological features, the meteorological features are divided into different intervals (table 1), meanwhile, whether non-meteorological features such as holidays (weekends), seasons, time periods and the like are introduced to carry out feature construction, frequent patterns between meteorological conditions and events are mined through an FP-Growth algorithm, event types with high correlation degree, large occurrence quantity and obvious increment with certain meteorological conditions are analyzed, potential rules existing in the meteorological conditions and the event data are mined according to the frequent patterns, and the credibility of the rules is quantized.

TABLE 1 Meteorological characteristic interval of automatic Meteorological observation station

Knowledge maps describe knowledge and objects and relationships among objects by using a graph structure, can express information in a form closer to human cognition, provide a capability of organizing, managing and understanding mass information, and are currently applied to a plurality of industry fields. According to the invention, by introducing a knowledge graph technology, mining results based on the FP-Growth frequent mode, and processing the subset superset, a knowledge graph rule base is completed, so that the link relation between weather and events can be better presented.

The results and analysis of the method of the invention are as follows:

1. event base feature analysis

The time series of grid-wide management events in the whole market generally show a trend of stable fluctuation (fig. 1), and the number of monthly events is between 60000 and 95000, wherein the number is the lowest in 1 month, 2 months and 10 months, and the highest in 8 months. The number of daily events is mostly in the interval from 15000 to 35000, wherein part of the dates have data missing, and the number of events is extremely low. As can be seen by fitting the trend lines, the number of events was generally higher between months 4-10, and lower between months 1-3 and 11-12.

The number distribution of events in one week is shown in fig. 2, and the period distribution is obvious from monday to sunday in a period of one week, and the difference between weekday and non-weekday is obvious, and the average daily event number on weekend is only 65% of weekday. The events were distributed mainly between 7 and 17, with a day period, with the number of events accounting for 95% of the total number of events, while events were seen to form two distinct peaks at 9 and 14, with valleys before and after 11 (fig. 3), which highly coincide with the daily work hours.

2. Event space feature analysis

The spatial distribution of the event density is highest in the central urban area, and the event density exceeds 20000/km ² Second, the inner-middle ring area immediately adjacent to the central urban area and the urban subcenter of each area, and it can be seen that the event density shows a tendency to decrease gradually from the central urban area to the suburban area. In general, areas of high event concentration all coincide with urban areas of high personnel concentration.

3. Event category feature analysis

The urban grid data has the characteristic of 'head concentration and long tail distribution' in categories, and after partial event types and subclasses with the number smaller than 1000 of the event types are removed from each area, the urban grid data totally comprises 31 major classes and 108 minor classes. And (4) conclusion: the proportion of the urban content and sanitation in the event categories exceeds 50 percent, wherein the quantities of exposed garbage and random scribbling, random posting and random carving are the largest, and respectively reach 32.0 percent and 17.1 percent of the total quantity of the screened events; the major categories of street order, garbage classification and road surface civilization supervision are ranked from the second to the fourth in number, the proportions are respectively 18.8%, 5.1% and 4.8%, and the corresponding major categories are motor vehicle casual parking, non-motor vehicle casual parking, residential district classification with poor effect and non-motor vehicle civilized traffic behavior.

The specific content of the event keyword and co-occurrence term analysis in the method of the invention is as follows:

1. grid patrol event keyword and word cloud analysis

A JiebaR Chinese word segmentation engine is adopted for word segmentation processing, and word cloud analysis is carried out on the basis of event keywords extracted by a TF-IDF algorithm. The event keywords are the highest in the occupation ratio of "garbage", "exposed", "posted" and "parked", and event descriptions such as "bicycle", "non-motor vehicle" are the second most, and the categories of the event descriptions such as exposed garbage, vehicle parking and advertisement posting are larger in the occupation ratio through word cloud analysis. Similarly, the invention carries out word cloud analysis on the major categories of events such as street order, pavement civilization supervision, public facilities and the like in sequence, takes the public facilities as an example, the words such as well covers, upright posts, loss and the like are higher in percentage, and the word cloud description is highly consistent with the category attribution.

2. Event co-occurrence term analysis

The term network graph (FIG. 4) may analyze the frequency of terms that occur in pairs in the event description and plot the relationship network. Some clear clusters can be seen from the lexical item network diagram of the event, such as the problem of parking disorderly (the terms related to "motor vehicle", "non-motor vehicle", "sharing", "single vehicle", "parking", etc.), the behavior of traffic civilization (the terms related to "riding", "wearing", "on the road", "helmet", "intersection", etc.), the problem of city appearance (the terms related to "city appearance", "messy coating", "advertisement", "posting", "airing", etc.), the problem of garbage throwing (the terms related to "garbage", "exposing", "living", "building", "street", "mixed throwing", "overflow", "squaring", etc.), the problem of municipal facility damage (such as "municipal", "vertical bar", "partition", "facility", "damage", "guardrail", etc.), the problem of rainwater well cover missing displacement (the terms related to "rainwater", "well cover", "grate", "missing", "displacement", "curbstone", etc.), and the event with obvious clustering characteristics also includes the problem of city management such as fire hydrant facility, street tree and cover plate damage, advertisement disorderly setting, green occupation of green, etc. Analysis shows that the keywords in the event description have strong co-occurrence conditions, a clear clustering structure can be formed, meanwhile, the clusters are concentrated around terms such as roads, residential buildings and the like, and a co-occurrence term relation network taking citizen activities as a main body is formed.

The specific contents of the weather and event rule mining in the method are as follows:

1. correlation analysis of weather conditions and events

On the basis of event analysis, the method and the system develop correlation analysis of meteorological conditions and events based on synchronous automatic meteorological observation data. In the research, the cumulative precipitation, the highest air temperature, the average air temperature, the lowest air temperature, the maximum wind speed and the average wind speed which are 6 hours, 12 hours and 24 hours before the occurrence of the event are respectively processed according to meteorological observation data, and are taken as characteristic quantities to count the average number of the events falling into different intervals in a segmented manner, and on the basis, the Pearson correlation coefficients of the grid subclass and different meteorological characteristics are respectively calculated. The results show that a certain correlation exists between part of events and meteorological conditions, the correlation between relevant subclasses of municipal facilities (including rainwater well covers, sewage well covers, fire hydrants, green guard rails, electric facilities (equipment) and the like) in the events and the temperature is obvious, and the conditions that the use and the maintenance of the relevant facilities and equipment are more frequent in summer and the damage or the non-return of the facilities are more advanced are shown; meanwhile, greening environmental sanitation events (including 'road cleaning', 'store along street classification actual effect', 'river pollution', 'cell greening' and the like) also show a high trend along with the rise of air temperature, which shows that the problems of greening cleaning, garbage classification and the like are more likely to occur along with the rise of air temperature. The wind-related events are mainly concentrated on wind-vulnerable structures such as vertical rods, flower stands, tree pit cover plates of street trees and the like (including traffic vertical rods, electric power rods, flower stand flower bowls and tree pit cover plate damage), and meanwhile, environmental events such as river pollution, river greening, community environment, agricultural waste disposal and the like are increased along with the increase of wind power. In addition, some illegal violations (such as "illegal handling of waste soil", "pedestrian uncivilized traffic behaviors", "scrawling, messy posting, random depiction", "shared vehicle is not parked in a specified area", "aerial toss", and the like) also have high correlation with weather conditions, indicating that the related events are more likely to occur under weather conditions of high temperature, rain, strong wind, and the like.

Under certain meteorological conditions, the number of partial events shows a relatively obvious downward trend, for example, the number of the 'random painting, the random posting and the random carving' is small in rainy days, and the number of the 'fire hydrant', the 'sewage well cover', 'road cleaning', 'random painting, the random posting and the random carving' and other events also tend to be reduced along with the increase of the wind speed.

2. Frequent pattern mining of meteorological conditions and typical events

A frequent pattern (frequency pattern) is a pattern that frequently appears in a data set (e.g., a set of items, a subsequence, or a substructure). According to the method, the meteorological conditions are divided into different intervals, the intervals are combined with the types of events occurring in the intervals, frequent pattern mining is carried out on the meteorological events and the events by adopting an FP-Growth algorithm, and a knowledge graph (shown in a figure 5 and a figure 6) of a typical event is constructed. The result shows that the meteorological conditions of the events are mostly calm weather without precipitation and with the average wind speed of 2-3 levels. In addition, the types of events such as traffic upright rods, electric power rods, river pollution, river greening, damage to tree pit cover plates, agricultural waste disposal, high-altitude throwing and the like are also frequently generated in windy days with the maximum wind speed of 4-5 grades; the events such as 'residential greening', 'river pollution', 'river greening', 'classification of shops along the street but not in place of practical effect', 'damage of tree pit cover plate', 'high altitude parabolic' and the like are mostly in an air temperature interval of 20-30 ℃. In general, the result of frequent pattern mining through FP-Growth is consistent with the correlation analysis conclusion.

Compared with a correlation analysis method, the FP-Growth can also be used for analyzing and mining the incident under the specific weather condition, and is beneficial to the advance handling and disposal aiming at the specific weather condition. According to FP-Growth excavation results, under the precipitation condition, events such as 'foundation pits', 'illegal occupation of underground public pedestrian paths', 'road surface water accumulation, sewage overflow, excrement overflow' and the like can be obtained (the occurring confidence and the support degree are 1.00, 0.41, 0.94, 0.33, 1.00 and 0.32 respectively); when the temperature is lower, events such as disputes, public fire safety hazards, market management and the like can be obtained (the occurring confidence and the support degree are 0.71, 0.42, 0.93, 0.40, 0.92 and 0.40 respectively; when the temperature is higher, events such as high altitude parabolic events, garbage in cells and the like can be obtained (the occurring confidence and the occurring support degree are 0.96, 0.41, 0.32, 0.86 and 0.31 respectively); when the wind speed is high, event types such as 'river greening', 'river pollution' and 'public service advertisement damage' can be obtained (the occurring confidence and the support degree are 0.85, 0.44, 0.88, 0.42, 0.92 and 0.41.

Based on the analysis, the invention takes traffic upright poles and river pollution as examples, constructs a typical event knowledge map covering meteorological conditions and urban operation situations, and forms a knowledge framework of triggering urban operation signs by the meteorological conditions by inducing and expressing the correlation between the meteorological conditions and the urban operation, thereby being beneficial to the advance prejudgment and the advance disposal of the urban operation situations under different meteorological conditions.

Based on R and Python language environments, 2021-year gridding management data are used as research data, space-time characteristic analysis, word cloud and co-occurrence term characteristic analysis, correlation analysis and frequent pattern mining methods are adopted to analyze and research city operation management big data, a typical event knowledge graph covering meteorological conditions is constructed, and the influence rule of the meteorological conditions on city operation is further disclosed. The main conclusions are as follows:

(1) The grid management event is highly correlated with the resident activity characteristics. The event occurrence time is highly consistent with the working time, the occurrence area is also consistent with the dense region of urban personnel, the phenomena of concentrated head and long tail distribution exist in the category, a clear clustering structure can be formed on the event participle, and a co-occurrence lexical relation network taking the citizen activities as the main body is formed.

(2) The weather conditions have strong correlation with part types of events, and under specific weather conditions, part of events show obvious high emergence situations. In the event, the relevance of the small categories such as municipal facilities, greening sanitation and the like to the air temperature is obvious, the wind-vulnerable structures such as the vertical rods are greatly influenced by the wind force, and illegal behaviors also have high relevance to meteorological conditions. In addition, partial events may present a clear and easy situation under certain weather conditions. The weather conditions of precipitation, low temperature, high temperature, strong wind and the like can be excavated to obtain the event types of 'foundation pits', 'disputes', 'river greening' and the like.

(3) The invention further introduces a knowledge graph technology, constructs a typical event knowledge graph covering meteorological conditions and urban operation situations, and forms a knowledge framework for triggering urban operation signs by the meteorological conditions by inducing and expressing the correlation between the meteorological conditions and the urban operation, thereby being beneficial to the advance prejudgment and the advance disposal of the urban operation situations under different meteorological conditions, and providing a certain decision reference for promoting fine management measures for cities and optimizing an urban management system.

The grid management data has a good indication significance for city operation signs, but the existing research has a great defect in revealing the time-space regularity of city management events on a fine scale, and weather conditions often affect the occurrence and development of the events, but the existing research is lack of the regular analysis of the weather conditions on the occurrence of the events. Therefore, the invention firstly adopts a space-time analysis and natural language processing method to carry out feature mining on the gridding event data, and on the basis, the influence analysis research of meteorological conditions on the occurrence of urban gridding management events is carried out by mainly combining meteorological observation data. In addition, the invention also introduces a frequent pattern mining algorithm FP-Growth, further carries out mining of association rules of meteorological conditions and event occurrence, and constructs a typical event knowledge graph covering the meteorological conditions. By identifying the correlation characteristics of weather and event data, the deep fusion of urban operation big data and weather big data can be realized, and technical support and decision reference are provided for urban accurate management.

The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. It will be apparent to those skilled in the art that various changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, and the scope of protection is still within the scope of the invention.

Claims (5)

1. A method for mining association rules of weather and city operation signs based on grid management is characterized by comprising the following steps:

step 1, data selection and source

The data information comprises urban grid patrol data and synchronous meteorological observation data, the grid patrol data is obtained in real time through a data interface provided by a building department, the meteorological observation data is from an automatic meteorological observation station distributed in the whole city range by a meteorological bureau, and four meteorological elements of precipitation, air temperature, wind direction and wind speed are selected to participate in analysis;

step 2, data processing

Under the python environment, converting the longitude and the latitude into a plane coordinate by adopting a universal transverse-axis mercator projection under a geodetic coordinate system, extracting events in a buffer area in a certain range around the automatic weather station as an analysis target, and carrying out weather element assignment on the events;

step 3, data research

Performing time-space characteristic analysis and type characteristic analysis of events under R and python language environments based on grid patrol data and meteorological observation data, and constructing a keyword cloud and a co-occurrence term network by adopting a Chinese word segmentation and keyword extraction method through event description information;

the time and the position of the event and the weather information are matched, the FP-Growth algorithm is utilized to mine the association rule of the weather condition and the typical event, the corresponding matching rule is obtained, and the graphical expression is carried out through a knowledge graph.

2. The grid management based weather and urban operation sign association rule mining method according to claim 1, wherein the Chinese word segmentation method comprises the following steps: selecting a JeibaR word segmentation packet in an R language environment to perform word segmentation processing on a grid description text, and performing the word segmentation processing by adopting a mixed word segmentation engine in combination with a maximum probability method and a hidden Markov model.

3. The grid management based weather and city operation sign association rule mining method according to claim 2, wherein the keyword extraction method comprises: on the basis of segmenting the grid description information, a TF-IDF algorithm is adopted to extract keywords corresponding to the events.

4. The grid management based weather and city operation sign association rule mining method according to claim 3, wherein the step of mining association rules of weather conditions and typical events by using FP-Growth algorithm comprises: selecting statistics of 6 hours, 12 hours and 24 hours of accumulated precipitation, the highest air temperature, the average air temperature, the lowest air temperature, the maximum wind speed and the average wind speed to construct meteorological features, dividing the meteorological features into different intervals, introducing non-meteorological features to perform feature construction, mining a frequent pattern between a meteorological condition and an occurred event through an FP-Growth algorithm, analyzing event types with high correlation degree with the meteorological condition, large occurrence quantity and obvious increment, mining potential rules in the meteorological condition and event data according to the frequent pattern, and quantifying the reliability of the rules.

5. The grid management based weather and urban operation sign association rule mining method according to claim 4, wherein the step of graphically expressing through a knowledge graph comprises: and mining results based on the FP-Growth frequent pattern, and processing the subset superset to complete a knowledge graph rule base and present a link relation between weather and events.

Images