CN104217087A - Permanent resident population analysis method based on operator network data - Google Patents

Abstract

The invention relates to a permanent resident population analysis method based on operator network data. According to the method, statistics is carried out on the network data of user crowds in a certain area on the basis of the collected operator network data by utilizing the main idea of an ant colony algorithm, classification is carried out on the data according to the appearance positions and time indexes of the user crowds, calculation is carried out on the results obtained through statistics and analysis the factors according to given functional transformation, the calculation result serves as a judging basis of population classification, iteration correction is carried out on the characteristic values (the permanent resident population) of the crowds, and new judging results are fed back. According to the permanent resident population analysis method based on the operator network data, the defects that an original statistical classification method is poor in regression, slow in convergence, abnormally sensitive to abnormal data, and prone to causing misjudgment and neglected judgment are overcome; the method has the advantages that the number of requested sample data is small, requirements for an initial value are not accurate, and correction can be carried out at any time according to data; the partition of the permanent resident population in the area can reflect the current situations more truly.

Description

A kind of permanent resident population's analytical approach based on carrier network data

Technical field

The invention belongs to data mining analysis technical field, be specifically related to a kind of permanent resident population's analytical approach based on carrier network data, it can judge the population classification in a certain region accurately, and namely which is permanent resident population, for relevant departments grasp timely, accurate, effective data.

Background technology

In current specific region, population situation is complicated, and movement of population is complicated, and national correlation department often only grasps static information, and these information are often delayed, upgrades not in time, can not provide good technical support for relevant Decision.Traditional permanent resident population's analytical approach often adopts fairly simple model, from concept intuitively, such as think that permanent resident population presents the fixing data in locality at night, and the mechanics on daytime is with regard to more complicated, gives up this situation and analyzes.Along with townie variation, this mechanics can not react present situation very accurately.In addition, the sample that traditional statistical method will obtain is enough large and abundant, and Return Difference, convergence slowly, and for abnormal data quite sensitive, easily cause erroneous judgement, fail to judge.

Ant group algorithm is a kind of probability type algorithm being used for finding in the drawings path optimizing.This algorithm, based on positive feedback principle, has that strong robustness, convergence are good, the advantage such as concurrency and adaptivity, is just used to the problem solving most of optimization problem or can be converted into Optimization Solution.Along with ant group algorithm development and perfect, its application has expanded to multiple-objection optimization, Data classification, data clusters, pattern-recognition, system modelling, flow layout, signal transacting, image procossing, decision support and the aspect such as emulation and System Identification at present.Ant group algorithm is suitable for the classification of user group's behavioural characteristic with its plurality of advantages very much.In addition, a large number of users data that operator grasps, can as the another kind of Data Source of real-time analysis population classification.

Summary of the invention

In view of this, the object of the invention is to for deficiency of the prior art, adopt ant group algorithm, in conjunction with operator's mass data, provide a kind of permanent resident population's analytical approach based on carrier network data, calculate listener clustering.

For completing above-mentioned purpose, the technical solution used in the present invention is: the main thought applying ant group algorithm on the basis of the carrier network data gathered, the network data of the user crowd in certain area is added up, classify according to its position occurred, time index, the result obtained by the above-mentioned factor statistical study of given function transform pairs carries out computing, the judging basis that it can be used as population to classify, thus iterated revision is carried out to the eigenwert (permanent resident population) of crowd, and return new evaluation result.Specifically, the invention provides a kind of permanent resident population's analytical approach based on carrier network data, comprise the following steps:

Steps A: the primary data in the Community Population information acquisition defined area that Census information, local police station grasp;

Step B: obtain historical data and real time data from operator, the content of data comprises collection number, acquisition time, collecting location;

Step C: classify according to position, time that user data occurs, and by given functional transformation, carrier data and primary data are iterated the matching analysis, thus raw data is divided into two large classes, one class can tentatively be determined, another kind of is the data of unusual fluctuation, namely because raw data lacks the new supplementary data caused;

Step D: for functional transformation analysis, introduce mechanics model, added to dynamically in knowledge network by these models, model comprises: peak period on and off duty rule model, period at night rule model, festivals or holidays rule model and weather conditional plan model;

Step e: according to the knowledge network enriched constantly, then according to the data of operator's Real-time Collection, make correction, by the information entropy concept of ant group algorithm, user is further classified.

Beneficial effect of the present invention is: 1. overcome original statistical classification method Return Difference, convergence is slow, and for abnormal data quite sensitive, easily causes erroneous judgement and the shortcoming of failing to judge; 2. the sample data with requirement is few, initial value requires out of true, can carry out the feature revised at any time, also can well introduce external factor simultaneously according to data, the external data typings such as such as Census information, can the corresponding weighted value of dynamic conditioning to different factors.

Accompanying drawing explanation

Below in conjunction with accompanying drawing, the present invention is further described.

Fig. 1 is the principle schematic of the permanent resident population's analytical approach that the present invention is based on carrier network data.

Embodiment

Be illustrated in figure 1 the principle schematic of the permanent resident population's analytical approach based on carrier network data provided by the invention, its analytical approach, comprises the following steps:

Steps A: the primary data in the Community Population information acquisition defined area that Census information, local police station grasp;

Step B: obtain historical data and real time data from operator, the content of data comprises collection number, acquisition time, collecting location; The granularity (being generally 2 hours) of data acquisition, when data are more, judging basis is more accurate;

Step C: classify according to position, time that user data occurs, and by given functional transformation, carrier data and primary data are iterated the matching analysis, thus raw data is divided into two large classes, one class can tentatively be determined, another kind of is the data of unusual fluctuation, namely because raw data lacks the new supplementary data caused;

Step D: for functional transformation analysis, introduce mechanics model, added to dynamically in knowledge network by these models, model comprises: peak period on and off duty rule model, period at night rule model, festivals or holidays rule model and weather conditional plan model;

Peak period on and off duty rule model: in this period, operator's image data frequency strengthens, it is generally acknowledged on and off duty outside daylight hours, permanent resident population's great majority are activities in one's respective area.

Period at night rule model: traditionally, be that differentiation is fairly simple in the evening, and because the activity work and rest rule of people shows, be all the time of having a rest in the evening, and the data of collection can well meet this feature.

Festivals or holidays rule model: this mechanics is the experience according to former years, during festivals or holidays, the routine plan of travel of permanent resident population.

Weather conditions rule model: the mechanics also affecting people when weather extremes occurs time, such as large rainy day people tend to unmovable in the street.

Step e: according to the knowledge network enriched constantly, then according to the data of operator's Real-time Collection, make correction, by the information entropy concept of ant group algorithm, user is further classified.

In above-mentioned steps, the initial value of functional transformation is foundation according to external data, and the proportion that these factors take constantly carries out revising according to the data in later stage.The model of functional transformation is also dynamic conditioning, is original external data constantly to be rejected, a process of refinement.

Contemplated by the invention discriminating and the analysis of bad data in data analysis process; Consider the several frequently seen factor affecting crowd activity's rule, such as work and rest on daytime, off-day, festivals or holidays, extreme weather; Consider shortage of data in certain period in time, system can carry out auto-complete according to knowledge network automatically, and revises this knowledge network according to during follow-up fructufy.Show by experiment, the method makes the division of the permanent resident population in region react present situation more really, and real-time have also been obtained good guarantee.

Above embodiment is to illustrate the invention and not to limit the present invention, and the equivalent variations that those skilled in the art carry out above-described embodiment, replacement are all within claims.

Claims (1)

1., based on permanent resident population's analytical approach of carrier network data, it is characterized in that: comprise the following steps:

Steps A: the primary data in the Community Population information acquisition defined area that Census information, local police station grasp;

Step B: obtain historical data and real time data from operator, the content of data comprises collection number, acquisition time, collecting location;

Step C: classify according to position, time that user data occurs, and by given functional transformation, carrier data and primary data are iterated the matching analysis, thus raw data is divided into two large classes, one class can tentatively be determined, another kind of is the data of unusual fluctuation, namely because raw data lacks the new supplementary data caused;

Step D: for functional transformation analysis, introduce mechanics model, added to dynamically in knowledge network by these models, model comprises: peak period on and off duty rule model, period at night rule model, festivals or holidays rule model and weather conditional plan model;

Step e: according to the knowledge network enriched constantly, then according to the data of operator's Real-time Collection, make correction, by the information entropy concept of ant group algorithm, user is further classified.