CN105183876A - Psychological pressure value predicting method and system based on microblog - Google Patents

Abstract

The invention provides a psychological pressure value predicting method and system based on microblog. The method includes the steps of detecting microblog information issued by a target object, obtaining the pressure value and issuing time of each piece of microblog information, establishing a first pressure-time sequence of the target object according to the pressure value and issuing time of each piece of microblog information, dividing the first pressure-time sequence according to the preset time granularity, obtaining a plurality of second pressure-time sequences corresponding to the preset time granularity, obtaining pressure values of microblog information in the second pressure-time sequences, establishing a time sequence model according to the pressure values of microblog information in each second pressure-time sequence, and determining the pressure value of the corresponding microblog information of the next second pressure-time sequence of the target object according to the time sequence model. The psychological pressure value of the target object can be predicted, so the target object can conveniently manage psychological pressure of himself/herself and can be prevented from potential harms caused by psychological pressure.

Description

A kind of psychological pressure value prediction method and system based on microblogging

Technical field

The present invention relates to Information Communication field, be specifically related to a kind of psychological pressure value prediction method and system based on microblogging.

Background technology

Adolescence is the important stage of a personal growth.In this Physiological Psychology fast-changing period, teenagers can be subjected to the pressure from each side, as school work pressure, and autognosis pressure, human communication pressure, and causalgia.Investigation display teenagers have become the colony that the U.S. has pressure most.1,950 adults and 1 have been interviewed in this investigation, 018 teenager, and in investigation, based on the pressure total score of 10 points, the pressure scoring of teenagers is 5.8, is greater than the pressure score value 5.1 of adult.The teenager of 31% represents the puzzlement that is deeply stressed, and the teenager of 30% to feel sorry for or dejected because of pressure.According to 19, Hong Kong middle school nearly 9, the investigation of 100 students finds: slight even serious depressive symptom appears in the student that is interviewed of 49.9%, as cryyed, disagreeable oneself, appetite change etc., their pressure is mainly from school work future, examination, the appearance bodily form etc.Exceeding twenty percent student once occurred slight to serious suicidal thought.

Above-mentioned investigation shows, teenagers are faced with huge pressure in their growth environment.Suppressing of continuing can all cause serious infringement to their physiology and mental health, and such as anxiety, hypertension, immunity degradation, depression, heart disease, even commit suiside; And due to the shortage of teenagers' experience and the unsound of psychological modulation mechanism, they may be unaware of the potential hazard of psychological pressure to oneself, to such an extent as to can not seek and effectively be helped.

Summary of the invention

For defect of the prior art, the invention provides a kind of psychological pressure value prediction method and system based on microblogging, destination object psychological pressure can be predicted in advance, so that destination object management pressure itself, and then the potential injury avoiding psychological pressure to bring.

The present invention proposes a kind of psychological pressure value prediction method based on microblogging, comprising:

The micro-blog information that destination object is issued is detected, obtains force value and the time of origin of every bar micro-blog information, and according to the force value of described every bar micro-blog information and time of origin, set up the first pressure time sequence of described destination object;

Described first pressure time sequence is divided according to Preset Time granularity, obtains the second pressure time sequence that multiple and described Preset Time granularity is corresponding;

Obtain the force value of micro-blog information in described each second pressure time sequence, according to force value series model Time Created of micro-blog information in described each second pressure time sequence, and according to the force value of described time series models determination destination object micro-blog information in next second pressure time sequence.

Optionally, the force value of micro-blog information in the described each second pressure time sequence of described acquisition, according to force value series model Time Created of micro-blog information in described each second pressure time sequence, comprising:

Gather maximal value, minimum value, the mean value of the force value of micro-blog information in each second pressure time sequence;

According to the accumulative of the maximal value of the force value of micro-blog information in all second pressure time sequences and force value and, set up described time series models;

Or

In all second pressure time sequences, the total number of the minimum value of the force value of micro-blog information, pressure microblogging ratio and microblogging, sets up described time series models;

Or

In all second pressure time sequences the mean value of the force value of micro-blog information, pressure microblogging ratio and force value accumulative and, set up described time series models;

Wherein, the Cumulate Sum of described force value be the force value of micro-blog information in all second pressure time sequences and accumulative and, the total number of described microblogging is the Cumulate Sum of the quantity of the micro-blog information issued in all second time serieses, described pressure microblogging ratio be in each second time series pressure microblogging ratio and Cumulate Sum, in described each second time series, pressure microblogging ratio is the quantity that the force value of micro-blog information in each second time series is greater than micro-blog information in the quantity of the micro-blog information of zero and each second time series.

Optionally, described time series models are:

$L_{n+1}=C+\underset{i=0}{\overset{k-1}{\Σ}}{A}_i{L}_{n-i}+\underset{i=0}{\overset{k-1}{\Σ}}{B}_i{X}_{n-i}+\underset{i=0}{\overset{k-1}{\Σ}}{\mathrm{\θ}}_i{\mathrm{\ϵ}}_{n-i}+{\mathrm{\ϵ}}_{n+1}$

Wherein, L _n+1for the force value of destination object micro-blog information in (n+1)th the second pressure time sequence, C is constant, A _i, B _iand θ _ifor parameter preset, k is the item number of model, L _n-k+1, L _n-k+2... L _nfor the force value of micro-blog information in front k the second pressure time sequence, ε _n-k+1, ε _n-k+2..., ε _n, ε _n+1be respectively and make E (ε _n-k+1)=0, E (ε _n-k+2)=0 ..., E (ε _n)=0, E (ε _n+1the white noise error item of)=0; X _n-k+1, X _n-k+2... X _nfor micro-blog information in front k the second pressure time sequence force value accumulative and, or be the accumulative of the force value of micro-blog information in front k the second pressure time sequence and and pressure microblogging ratio, or be the pressure microblogging ratio of force value and the number of micro-blog information of micro-blog information in the individual second pressure time sequence of front k.

Optionally, after according to the force value of described time series models determination destination object micro-blog information in next second pressure time sequence, comprising:

According to force value and the corresponding pressure events of micro-blog information in described each second pressure time sequence, build-up pressure changing pattern collection model;

The micro-blog information issued according to described destination object and the force value of micro-blog information, obtain the initial time of goal pressure event corresponding to micro-blog information and described goal pressure event;

According to the type of described goal pressure event, described goal pressure event is mated with described pressure changing pattern collection model, obtains the pressure changing pattern that described goal pressure event is corresponding;

According to the initial time of described goal pressure event, by the pressure changing pattern of described goal pressure event, obtain the modified value of described goal pressure event to described destination object force value of micro-blog information in next second pressure time sequence;

According to the modified value of described goal pressure event to described destination object force value of micro-blog information in next second pressure time sequence, the force value of destination object micro-blog information in next second pressure time sequence that described time series models are determined is revised, obtains the forecast pressure value of described destination object micro-blog information in next second pressure time sequence.

Optionally, described according to the modified value of described goal pressure event to destination object force value of micro-blog information in next second pressure time sequence, the force value of destination object micro-blog information in next second pressure time sequence that described time series models are determined is revised, comprising:

The pressure changing pattern corresponding to described goal pressure event divides according to event evolves period, obtains the modified value of described pressure changing pattern at different times;

The micro-blog information corresponding according to described goal pressure event, obtains the developing stage of described goal pressure event;

The different times of described pressure changing pattern is mated with the developing stage of described goal pressure event, obtains the modified value matched with the developing stage of described goal pressure event;

According to the modified value that the developing stage of described and described goal pressure event matches, superposition correction is carried out to the force value of destination object micro-blog information in next second pressure time sequence that described time series models are determined.

Present invention also offers a kind of psychological pressure value prediction system based on microblogging, comprising:

Detection module, detects for the micro-blog information issued destination object, obtains force value and the time of origin of every bar micro-blog information;

First sets up module, for according to the force value of described every bar micro-blog information and time of origin, sets up the first pressure time sequence of described destination object;

Dividing module, for dividing according to Preset Time granularity described first pressure time sequence, obtaining the second pressure time sequence that multiple and described Preset Time granularity is corresponding;

Second sets up module, for obtaining the force value of micro-blog information in described each second pressure time sequence, and according to force value series model Time Created of micro-blog information in described each second pressure time sequence;

Determination module, for the force value of micro-blog information of answering at next second pressure time sequence pair according to described time series models determination destination object.

Optionally, second set up module for:

Gather maximal value, minimum value, the mean value of the force value of micro-blog information in each second pressure time sequence;

According to the accumulative of the maximal value of the force value of micro-blog information in all second pressure time sequences and force value and, set up described time series models;

Or

In all second pressure time sequences, the total number of the minimum value of the force value of micro-blog information, pressure microblogging ratio and microblogging, sets up described time series models;

Or

In all second pressure time sequences the mean value of the force value of micro-blog information, pressure microblogging ratio and force value accumulative and, set up described time series models;

Wherein, the Cumulate Sum of described force value be the force value of micro-blog information in all second pressure time sequences and accumulative and, the total number of described microblogging is the Cumulate Sum of the quantity of the micro-blog information issued in all second time serieses, described pressure microblogging ratio be in each second time series pressure microblogging ratio and Cumulate Sum, in described each second time series, pressure microblogging ratio is the quantity that the force value of micro-blog information in each second time series is greater than micro-blog information in the quantity of the micro-blog information of zero and each second time series.

Optionally, described time series models are:

$L_{n+1}=C+\underset{i=0}{\overset{k-1}{\Σ}}{A}_i{L}_{n-i}+\underset{i=0}{\overset{k-1}{\Σ}}{B}_i{X}_{n-i}+\underset{i=0}{\overset{k-1}{\Σ}}{\mathrm{\θ}}_i{\mathrm{\ϵ}}_{n-i}+{\mathrm{\ϵ}}_{n+1}$

Wherein, L _n+1for the force value of the micro-blog information that destination object is answered at next second pressure time sequence pair, C is constant, L _n-k+1, L _n-k+2... L _nfor the force value of micro-blog information in front k the second pressure time sequence, A _i, B _iand θ _ifor parameter preset, k is the order of model, ε _ifor making E (ε _ithe white noise error item of)=0; X _n-k+1, X _n-k+2... X _nfor micro-blog information in front k the second pressure time sequence force value accumulative and, or, for micro-blog information in front k the second pressure time sequence force value accumulative and and pressure microblogging ratio; Or, be the pressure microblogging ratio of force value and the number of micro-blog information of micro-blog information in front k the second pressure time sequence.

Optionally, this system also comprises:

Correcting module, for according to the force value of micro-blog information in described each second pressure time sequence and corresponding pressure events, build-up pressure changing pattern collection model;

The micro-blog information issued according to described destination object and the force value of micro-blog information, obtain the initial time of goal pressure event corresponding to micro-blog information and described goal pressure event;

According to the type of described goal pressure event, described goal pressure event is mated with described pressure changing pattern collection model, obtains the pressure changing pattern that described goal pressure event is corresponding;

According to the initial time of described goal pressure event, by the pressure changing pattern of described goal pressure event, obtain the modified value of described goal pressure event to described destination object force value of micro-blog information in next second pressure time sequence;

According to the modified value of described goal pressure event to described destination object force value of micro-blog information in next second pressure time sequence, the force value of destination object micro-blog information in next second pressure time sequence that described time series models are determined is revised, obtains the forecast pressure value of described destination object micro-blog information in next second pressure time sequence.

Optionally, described correcting module is used for:

The pressure changing pattern corresponding to described goal pressure event divided according to the developing period of event, obtained the modified value of described pressure changing pattern at different times;

The micro-blog information corresponding according to described goal pressure event, obtains the developing stage of described goal pressure event;

The different times of described pressure changing pattern is mated with the developing stage of described goal pressure event, obtains the modified value matched with the developing stage of described goal pressure event;

According to the modified value that the developing stage of described and described goal pressure event matches, superposition correction is carried out to the force value of destination object micro-blog information in next second pressure time sequence that described time series models are determined.

The psychological pressure value prediction method based on microblogging that the present invention proposes, the force value of the every bar micro-blog information issued by destination object and time of origin, set up the first pressure time sequence about destination object, and according to the time granularity preset, first pressure time sequence is divided into multiple second pressure time sequence, and by force value series model Time Created of micro-blog information in each second pressure time sequence, with the force value of the micro-blog information of answering at next second pressure time sequence pair according to time series models determination destination object; The force value of next the second pressure time sequence of energy target of prediction object of the present invention, and then obtain the degree of pressure in destination object future, so that destination object takes healthy positive method, help oneself relief pressure, and then the potential injury avoiding psychological pressure to bring.

Accompanying drawing explanation

Can understanding the features and advantages of the present invention clearly by reference to accompanying drawing, accompanying drawing is schematic and should not be construed as and carry out any restriction to the present invention, in the accompanying drawings:

Fig. 1 shows the schematic flow sheet of the psychological pressure value prediction method based on microblogging that one embodiment of the invention provides;

Fig. 2 shows the destination object pressure mean values sequence based on all granularities of the psychological pressure value prediction method based on microblogging that one embodiment of the invention provides;

Fig. 3 shows the destination object pressure mean values sequence based on moon granularity of the psychological pressure value prediction method based on microblogging that one embodiment of the invention provides;

Fig. 4 shows the representative pattern of the school work types of events of the psychological pressure value prediction method based on microblogging that one embodiment of the invention provides.

Fig. 5 shows the structural representation of the psychological pressure value prediction system based on microblogging that one embodiment of the invention provides;

Fig. 6 shows the structural representation of the psychological pressure value prediction system based on microblogging that another embodiment of the present invention provides.

Embodiment

For making the object of the embodiment of the present invention, technical scheme and advantage clearly, below in conjunction with the accompanying drawing in the embodiment of the present invention, technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is a part of embodiment of the present invention, instead of whole embodiments.Based on the embodiment in the present invention, the every other embodiment that those of ordinary skill in the art obtain under the prerequisite not making creative work, all belongs to the scope of protection of the invention.

The schematic flow sheet of the psychological pressure value prediction method based on microblogging that Fig. 1 provides for one embodiment of the invention, with reference to figure 1, the method comprises the following steps:

Step 101: the micro-blog information that destination object is issued is detected, obtains force value and the time of origin of every bar micro-blog information, and according to the force value of every bar micro-blog information and time of origin, set up the first pressure time sequence of destination object;

Step 102: divide according to Preset Time granularity the first pressure time sequence, obtains multiple second pressure time sequence corresponding with Preset Time granularity;

Step 103: the force value obtaining micro-blog information in each second pressure time sequence, according to force value series model Time Created of micro-blog information in each second pressure time sequence, and the force value of the micro-blog information of answering at next second pressure time sequence pair according to time series models determination destination object.

The present invention propose based on microblogging psychological pressure value prediction method, the force value of the every bar micro-blog information issued by destination object and time of origin, set up the first pressure time sequence about destination object, and according to the time granularity preset, first pressure time sequence is divided into multiple second pressure time sequence, and by force value series model Time Created of micro-blog information in each second pressure time sequence, with the force value of the micro-blog information of answering at next second pressure time sequence pair according to time series models determination destination object; The force value of next the second pressure time sequence of energy target of prediction object of the present invention, and then obtain the degree of pressure in destination object future, so that destination object takes healthy positive method, help oneself relief pressure, and then the potential injury avoiding psychological pressure to bring.

In order to realize the prediction to maximal value, minimum value and the mean value in the force value of micro-blog information corresponding in next second pressure time sequence of destination object, the present invention establishes time series models.Concrete steps are as follows:

Step 131: the present invention acquires maximal value, minimum value, the mean value of the force value of micro-blog information in each second pressure time sequence.

Step 132: according to the maximal value of micro-blog information corresponding to the force value of micro-blog information in all second pressure time sequences and the accumulative of force value and, set up the time series models of the maximal value being used for the micro-blog information that force value is corresponding in next second pressure time sequence of target of prediction object.

Step 133: according to the total number of the minimum value of the force value of micro-blog information in all second pressure time sequences, pressure microblogging ratio and microblogging, sets up the time series models of the minimum value being used for the micro-blog information that force value is corresponding in next second pressure time sequence of target of prediction object.

Step 134: according to the accumulative of the mean value of the force value of micro-blog information in all second pressure time sequences, pressure microblogging ratio and force value and, set up the time series models of the mean value being used for the micro-blog information that force value is corresponding in next second pressure time sequence of target of prediction object.

It should be noted that, the Cumulate Sum of force value be the force value of micro-blog information in all second pressure time sequences and accumulative and, the total number of microblogging is the Cumulate Sum of the quantity of the micro-blog information issued in all second time serieses, pressure microblogging ratio be in each second time series pressure microblogging ratio and Cumulate Sum, in each second time series, pressure microblogging ratio is the quantity that the force value of micro-blog information in each second time series is greater than micro-blog information in the quantity of the micro-blog information of zero and each second time series.

Time series models of the present invention, by introducing multiple factor in forecasting process, next second seasonal effect in time series force value of the accurate target of prediction object of energy, compared with the time series models considering single factor with existing, the predicated error of multifactorial time series models of the present invention is all less than monofactorial time series models in multiple predicated error index, experimental data shows, time series models predicated error ratio of the present invention latter reduces 34%.

Below for the teenager in target group, the process of establishing of time series models is described in detail:

The first step, make I=[I.s, I.e] be a second pressure time sequence, wherein I.s represents the start time, and I.e represents the end time.The length of the second pressure time sequence is | I|=I.e-I.s, it can be one day, one week, January etc.Make W (I)=< (t ₁, w ₁), (t ₂, w ₂) ..., (t _m, w _m) > is micro-blog information sequence in I the second pressure time sequence, wherein w ₁, w ₂... w _nfor destination object is at time point t ₁, t ₂... t _nmicro-blog information (I.s≤the t issued ₁≤ ... ≤ t _m≤ I.e).

For each micro-blog information in W (I), by analyzing its series of features, such as: the content of text of microblogging comprises the number of the number of negative emotions vocabulary, front and negative emoticon, exclamation mark and question mark, the frequency of the picture shared or music and issuing microblog and time, and the degree of pressure being carried out each micro-blog information of perception by the method for machine learning.Adopt Stress (t _i, w _i)=(t _i, l _i) represent the pressure result of detection, wherein t _irepresent the time that microblogging is issued, l _irepresent the force value that this micro-blog information embodies, l _ican value 0 ~ 6, correspond respectively to and there is no pressure, slight pressure, less pressure, intermediate pressure, heavier pressure and the different degree of pressure of severe stress 6 kinds.Based on this, for a microblogging sequence W (I), obtain a force value sequence based on wall scroll microblogging:

S(W(I))＝＜Stree(t ₁,w ₁),Stree(t ₂,w ₂),...Stree(t _n,w _n)＞

Second step, according to the time granularity (month preset, one week or one day etc.) and prediction index (mean value, minimum value and maximum pressure value etc.), the present invention gathers together each for destination object the second pressure time sequence with three aggregate functions (Avg, Min and Max).

Three aggregate functions are as follows:

$Avg\left(S\right(W\left(I\right)\left)\right)=\frac{{\&Sigma;}_{i=1}^m{l}_i}{m}$

Min(S(W(I)))＝min(l _i) _1≤i≤m

Max(S(W(I)))＝max(l _i) _1≤i≤m

Wherein, m is the quantity of the microblogging of a second pressure time sequence internal object object publishing; Avg (S (W (I))), Min (S (W (I))) and Max (S (W (I))) are illustrated respectively in the mean value of the force value detected in a second pressure time sequence, minimum value and maximal value.

3rd step, in order to improve the degree of accuracy of prediction, the application also to the force value with prediction have the accumulative of user's pressure of statistic correlation and, the microblogging sent out total, the ratio that the number of pressure microblogging and pressure microblogging account for total microblogging number is assembled.

The present invention uses function Sum in I the second pressure time sequence, Scount and Ratio have accumulated above four indexs respectively, and these four aggregate functions are respectively:

$Sum\left(S\right(W\left(I\right)\left)\right)=\underset{i=1}{\overset{m}{\&Sigma;}}{l}_i$

Count(S(W(I)))＝m

Scount(S(W(I)))＝|{s _i|(1≤i≤m)^(l _i≠0)}|

Ratio(S(W(I)))＝Scount(S(W(I)))/Count(S(W(I)))

The gathering result of above-mentioned 7 aggregate functions in I the second pressure time sequence is expressed as L _{i, avg}, L _{i, min}, L _{i, max}, L _{i, sum}, L _{i, scount}, L _{i, count}and L _{i, ratio};

4th step, by assembling in n continuous print second pressure time sequence, obtaining 7 and assemble sequence, is L respectively _avg, L _min, L _max, L _sum, L _scount, L _countand L _ratio.

The present invention is according at the second pressure time sequence I ₁, I ₂, I ₃i _nresult (i.e., the L of interior build-up of pressure function _avg, L _min, L _max, L _sum, L _scount, L _countand L _ratio), to predict maximal value, minimum value and the mean value in the force value of micro-blog information corresponding in next second pressure time sequence of young people, and then obtain the degree of pressure of teenager's future time.

Fig. 2 and Fig. 3 be respectively the psychological pressure value prediction method based on microblogging that one embodiment of the invention provides all granularities and the moon granularity destination object pressure mean values sequence, referring to figs. 2 and 3, the pressure time sequence of young people presents seasonal pattern.The feature of this pattern is that pressure is smaller when vacation, comparatively large in pressure ratio in the test period, and the present invention adopts seasonal autoregressive model to solve nodositas characteristic and the instability of data in conjunction with the method for moving average time sequential analysis.

The pressure at teenager user lower a moment average, minimum, average, minimum not only with past pressure of maximal value, maximal value is relevant, and with the sum of issuing microblog, the ratio of pressure microblogging number and pressure microblogging is relevant.The Granger causality analytic approach (Grangercausalityanalysis) that the present invention uses determines the correlationship between each factor.

Its ultimate principle is: if predicting the outcome of obtaining of the sequential value of service factor Y obtains predicting the outcome good than using all factors but not comprising factor Y sequential value, then thinking that factor Y is granger cause, is useful.

Based on the degree of confidence of 95%, the mean value L of test result display pressure _avgwith pressure accumulated and be worth L _sumwith pressure microblogging ratio L _ratiorelevant; Pressure maximum value L _maxadd up with pressure and be worth L _sumrelevant; Pressure minimum L _minwith pressure microblogging ratio L _ratiowith issuing microblog total number L _countrelevant.Therefore, the next one using above-mentioned three correlative factor collection to predict as feature set is average, minimum, maximum pressure value.

Specifically, for the L that predicts the outcome of subsequent time n+1 _n+1(L _n+1∈ { L _{n+1, avg}, L _{n+1, max}, L _{n+1, avg}), the present invention considers the time sequential value (X with predicted target values correlative factor in past k chronomere _n-k+1, X _n-k+2, X _n-k+3x _n) wherein vectorial X _ifor the feature set of correlative factor.Such as, the present invention is by (L _{i, sum}, L _{i, ratio}) predict L as feature set _{i, avg}, L _{i, sum}l is predicted as feature set _{i, max}, (L _{i, count}, L _{i, ratio}) predict L as feature set _{i, min}.Its formula is as follows:

$L_{n+1}=C+\underset{i=0}{\overset{k-1}{\&Sigma;}}{A}_i{L}_{n-i}+\underset{i=0}{\overset{k-1}{\&Sigma;}}{B}_i{X}_{n-i}+\underset{i=0}{\overset{k-1}{\&Sigma;}}{\mathrm{\&theta;}}_i{\mathrm{\&epsiv;}}_{n-i}+{\mathrm{\&epsiv;}}_{n+1}$

Wherein, L _n+1for the force value of the micro-blog information that destination object is answered at next second pressure time sequence pair, C is constant, L _n-k+1, L _n-k+2... L _nfor the force value of micro-blog information in front k the second pressure time sequence, A _i, B _iand θ _ifor parameter preset, k is the order of model, ε _ifor making E (ε _ithe white noise error item of)=0; X _n-k+1, X _n-k+2... X _nfor micro-blog information in front k the second pressure time sequence force value accumulative and, or, for micro-blog information in front k the second pressure time sequence force value accumulative and and pressure microblogging ratio; Or, be the pressure microblogging ratio of force value and the number of micro-blog information of micro-blog information in front k the second pressure time sequence.

Better predict the outcome to obtain, the present invention utilizes AIC model (Akaike ' sInformationCriterion) to determine the value of k in time series models.AIC rule provides a kind of method weighing fitting effect and model complexity, it also selects suitable parameter value by AIC=2*N-2*ln (L) based on maximal possibility estimation, wherein N is the number needing estimated parameter, and L is the maximal value of likelihood function.When AIC value is less, represent the better of model and data fitting.According to minimum AIC value, obtain most suitable k value, and the k value that use obtains is used as the preset parameter in forecast model training.

The representative pattern of the school work types of events of the psychological pressure value prediction method based on microblogging that Fig. 4 provides for one embodiment of the invention, with reference to figure 4, in order to improve the degree of accuracy predicted the outcome, on the result that the present invention predicts at back, the result that the impact further combined with pressure events adjusts back prediction predicts the outcome more accurately to obtain.

The present invention after according to the force value of time series models determination destination object micro-blog information in next second pressure time sequence, also comprises:

401, according to the force value of micro-blog information and the pressure events of correspondence in each second pressure time sequence, build-up pressure changing pattern collection model;

402, according to the micro-blog information of destination object issue and the force value of micro-blog information, the initial time of goal pressure event corresponding to micro-blog information and goal pressure event is obtained;

403, according to the type of goal pressure event, goal pressure event is mated with pressure changing pattern collection model, obtain the pressure changing pattern that goal pressure event is corresponding;

404, according to the initial time of goal pressure event, by the pressure changing pattern of goal pressure event, goal pressure event is obtained to the modified value of destination object force value of micro-blog information in next second pressure time sequence;

405, according to the modified value of goal pressure event to destination object force value of micro-blog information in next second pressure time sequence, the force value of destination object micro-blog information in next second pressure time sequence that time series models are determined is revised, obtains the forecast pressure value of destination object micro-blog information in next second pressure time sequence.

Wherein, revise according to the force value of modified value to destination object micro-blog information in next second pressure time sequence that time series models are determined, specifically comprise:

The pressure changing pattern corresponding to goal pressure event divided according to the developing period of event, obtained the modified value of pressure changing pattern at different times;

The micro-blog information corresponding according to goal pressure event, obtains the developing stage of goal pressure event;

The different times of pressure changing pattern is mated with the developing stage of goal pressure event, obtains the modified value matched with the developing stage of goal pressure event;

According to the modified value that the developing stage with goal pressure event matches, superposition correction is carried out to the force value of destination object micro-blog information in next second pressure time sequence that time series models are determined.

Process impact in conjunction with pressure events being adjusted to the result of back prediction is below illustrated, concrete:

Assuming that I ₁, I ₂, I ₃... I _nbe n continuous print second pressure time sequence in a periods of events, wherein n represents the length of this goal pressure periods of events.Adopt Tr (I _x) represent the pressure trend of a unit interval, and adopt Tr=[Tr (I ₁), Tr (I ₂) ... Tr (I _n)] represent the impact of whole pressure events on destination object pressure, wherein, Tr (I _x) represent with exponential smoothing similarities and differences average line MACD (MovingAverageConvergence/Divergence), the variation tendency of destination object pressure under this events affecting is namely represented by the MACD value of pressure.MACD is a trend indicator being widely used in Analysis of Price of Stock, if x > is y, its formula is:

$MACD(x,y)=\frac{{\&Sigma;}_{j=x-y+1}^x{L}_j}{y}$

Wherein, x represents an xth moment, and y represents the size of time window, L _jrepresent the force value of moment j.

If x < is y, then formula is:

$MACD(x,y)=\frac{{\&Sigma;}_{j=1}^x{L}_j}{x}$

Based on MACD index, the present invention Tr=[Tr (I ₁), Tr (I ₂) ... Tr (I _n)] portray the pressure influence of event to destination object, wherein:

Tr(I _x)＝MACD(x,y ₁)-MACD(x,y ₂)y ₁＜y ₂,x＝1,2,3…n

In forecast model of the present invention, the just force value in prediction next moment, therefore by time window y ₁be set to 1, y ₂be 2.0 ~ 5 therefore the degree of pressure value of destination object is 5 to the maximum, and minimum is 0, then corresponding MACD value interval is, Tr (I _x) value to be scope be-5 ~+5 successive value.In order to carry out next step Frequent Pattern Mining, by Tr (I _x) span is divided into 40 sub-ranges: [0 ~ 0.25], [0.25 ~ 0.5] ..., [4.75 ~ 5] and [-0.25 ~ 0], [-0.5 ~-0.25], [-5 ~-4.75] are also mapped to corresponding integer label 1,2 ... 20, and-1 ,-2 ...-20.

By to the observation of teenager's microblog data and statistics, above-mentioned event is divided into school work pressure events and the large type of causalgia event two.Based on the microblogging of user, located the time period of a series of goal pressure event, but beginning and the Close Date of accurate positioning time is only difficult to according to content of microblog, therefore, the present invention supposes that the developmental sequence of event is one by one, known, the cycle that affects of an event starts from the time point of pressure from 0 rising, ends at the time point that force value reduces to 0.Based on above hypothesis, the present invention can locate a series of pressure events cycle, and calculates the MACD sequence representing the impact of this pressure events.According to the type of pressure events, the MACD sequence obtained is divided into school work pressure changing pattern collection and causalgia changing pattern collection by the present invention.For each set of patterns, GSP (GeneralizedSequentialPattern) algorithm is adopted to go to excavate the impact that changes the pressure of user to represent the type event of the longest Frequent Sequential Patterns that degree of confidence is 50%.

In last prediction, the present invention regards goal pressure event as a kind of overlaying influence, be about to the representative pattern out excavated be added to time series models prediction result on predict the outcome as last.As shown in Figure 4, the present invention will excavation representative pattern Tr out ^'early stage is divided into according to time average mid-term Tr ^'(I _{n/3 ~ 2n/3}) and later stage Tr ^'(I _{2n/3 ~ n}) three phases.And with (E _early, E _middle, E _later) portray that pressure events three phases produces added influence, wherein E _early, E _middleand E _laterbe respectively and Tr ^'(I _{2n/3 ~ n}) mean value of sequence.

First the present invention obtains a predicted value by time series models, then according to the extra information about goal pressure event, such as lower a moment power event with no pressure, lower a moment is by generation pressure events, pressure events is occurring and pressure events closes to an end, respectively by 0, E _early, E _middleand E _laterthe initial predicted that is added to result predicts the outcome as final.In order to the effect of measurement event correction, the present invention compares predicting the outcome of event correction and predicting the outcome without event correction further.Reach a conclusion: by the correction to event pressure influence, predict the error-reduction of future pressure degree value 18%.

The structural representation of the psychological pressure value prediction system based on microblogging that Fig. 5 provides for one embodiment of the invention, with reference to Fig. 5, this system comprises:

Detection module 51, detects for the micro-blog information issued destination object, obtains force value and the time of origin of every bar micro-blog information;

First sets up module 52, for according to the force value of every bar micro-blog information and time of origin, sets up the first pressure time sequence of destination object;

Dividing module 53, for dividing according to Preset Time granularity the first pressure time sequence, obtaining multiple second pressure time sequence corresponding with Preset Time granularity;

Second sets up module 54, for obtaining the force value of micro-blog information in each second pressure time sequence, and according to force value series model Time Created of micro-blog information in each second pressure time sequence;

Determination module 55, for the force value of micro-blog information of answering at next second pressure time sequence pair according to time series models determination destination object.

The present invention propose based on microblogging psychological pressure value prediction system, the force value of the every bar micro-blog information issued by destination object and time of origin, set up the first pressure time sequence about destination object, and according to the time granularity preset, first pressure time sequence is divided into multiple second pressure time sequence, and by force value series model Time Created of micro-blog information in each second pressure time sequence, with the force value of the micro-blog information of answering at next second pressure time sequence pair according to time series models determination destination object; The force value of next the second pressure time sequence of energy target of prediction object of the present invention, and then obtain the degree of pressure in destination object future, so that destination object takes healthy positive method, help oneself relief pressure, and then the potential injury avoiding psychological pressure to bring.

Wherein, second module 54 is set up for the maximal value, minimum value, the mean value that gather the force value of micro-blog information in each second pressure time sequence;

According to the accumulative of the maximal value of the force value of micro-blog information in all second pressure time sequences and force value and, Time Created series model;

Or the total number of the minimum value of the force value of micro-blog information, pressure microblogging ratio and microblogging in all second pressure time sequences, Time Created series model;

Or in all second pressure time sequences the mean value of the force value of micro-blog information, pressure microblogging ratio and force value accumulative and, Time Created series model;

Wherein, the Cumulate Sum of force value be the force value of micro-blog information in all second pressure time sequences and accumulative and, the total number of microblogging is the Cumulate Sum of the quantity of the micro-blog information issued in all second time serieses, pressure microblogging ratio be in each second time series pressure microblogging ratio and Cumulate Sum, in each second time series, pressure microblogging ratio is the quantity that the force value of micro-blog information in each second time series is greater than micro-blog information in the quantity of the micro-blog information of zero and each second time series.

Time series models are:

$L_{n+1}=C+\underset{i=0}{\overset{k-1}{\&Sigma;}}{A}_i{L}_{n-i}+\underset{i=0}{\overset{k-1}{\&Sigma;}}{B}_i{X}_{n-i}+\underset{i=0}{\overset{k-1}{\&Sigma;}}{\mathrm{\&theta;}}_i{\mathrm{\&epsiv;}}_{n-i}+{\mathrm{\&epsiv;}}_{n+1}$

Wherein, L _n+1for the force value of the micro-blog information that destination object is answered at next second pressure time sequence pair, C is constant, L _n-k+1, L _n-k+2... L _nfor the force value of micro-blog information in front k the second pressure time sequence, A _i, B _iand θ _ifor parameter preset, k is the order of model, ε _ifor making E (ε _ithe white noise error item of)=0; X _n-k+1, X _n-k+2... X _nfor micro-blog information in front k the second pressure time sequence force value accumulative and, or, for micro-blog information in front k the second pressure time sequence force value accumulative and and pressure microblogging ratio; Or, be the pressure microblogging ratio of force value and the number of micro-blog information of micro-blog information in front k the second pressure time sequence.

The structural representation of the psychological pressure value prediction system based on microblogging that Fig. 6 provides for another embodiment of the present invention, with reference to figure 6, in order to improve the degree of accuracy predicted the outcome, on the result that the present invention predicts at back, the result that the impact further combined with goal pressure event adjusts back prediction predicts the outcome more accurately to obtain.

The present invention is provided with correcting module, for according to the force value of micro-blog information in each second pressure time sequence and corresponding pressure events, and build-up pressure changing pattern collection model;

The micro-blog information issued according to destination object and the force value of micro-blog information, obtain the initial time of goal pressure event corresponding to micro-blog information and goal pressure event;

According to the type of goal pressure event, goal pressure event is mated with pressure changing pattern collection model, obtain the pressure changing pattern that goal pressure event is corresponding;

According to the initial time of goal pressure event, by the pressure changing pattern of goal pressure event, obtain goal pressure event to the modified value of destination object force value of micro-blog information in next second pressure time sequence;

According to the modified value of goal pressure event to destination object force value of micro-blog information in next second pressure time sequence, the force value of destination object micro-blog information in next second pressure time sequence that time series models are determined is revised, obtains the forecast pressure value of destination object micro-blog information in next second pressure time sequence.

Because this system and said method are mutually corresponding, therefore, repeat no more herein.

Although describe embodiments of the present invention by reference to the accompanying drawings, but those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, such amendment and modification all fall into by within claims limited range.

Claims (10)

1., based on a psychological pressure value prediction method for microblogging, it is characterized in that, comprising:

The micro-blog information that destination object is issued is detected, obtains force value and the time of origin of every bar micro-blog information, and according to the force value of described every bar micro-blog information and time of origin, set up the first pressure time sequence of described destination object;

Described first pressure time sequence is divided according to Preset Time granularity, obtains the second pressure time sequence that multiple and described Preset Time granularity is corresponding;

Obtain the force value of micro-blog information in described each second pressure time sequence, according to force value series model Time Created of micro-blog information in described each second pressure time sequence, and according to the force value of described time series models determination destination object micro-blog information in next second pressure time sequence.

2. method according to claim 1, is characterized in that, the force value of micro-blog information in the described each second pressure time sequence of described acquisition, according to force value series model Time Created of micro-blog information in described each second pressure time sequence, comprising:

Gather maximal value, minimum value, the mean value of the force value of micro-blog information in each second pressure time sequence;

According to the accumulative of the maximal value of the force value of micro-blog information in all second pressure time sequences and force value and, set up described time series models;

Or

In all second pressure time sequences, the total number of the minimum value of the force value of micro-blog information, pressure microblogging ratio and microblogging, sets up described time series models;

Or

In all second pressure time sequences the mean value of the force value of micro-blog information, pressure microblogging ratio and force value accumulative and, set up described time series models;

Wherein, the Cumulate Sum of described force value be the force value of micro-blog information in all second pressure time sequences and accumulative and, the total number of described microblogging is the Cumulate Sum of the quantity of the micro-blog information issued in all second time serieses, described pressure microblogging ratio be in each second time series pressure microblogging ratio and Cumulate Sum, in described each second time series, pressure microblogging ratio is the quantity that the force value of micro-blog information in each second time series is greater than micro-blog information in the quantity of the micro-blog information of zero and each second time series.

3. method according to claim 2, is characterized in that, described time series models are:

$L_{n+1}=C+\underset{i=0}{\overset{k-1}{\&Sigma;}}{A}_i{L}_{n-i}+\underset{i=0}{\overset{k-1}{\&Sigma;}}{B}_i{X}_{n-i}+\underset{i=0}{\overset{k-1}{\&Sigma;}}{\mathrm{\&theta;}}_i{\mathrm{\&epsiv;}}_{n-i}+{\mathrm{\&epsiv;}}_{n+1}$

Wherein, L _n+1for the force value of destination object micro-blog information in (n+1)th the second pressure time sequence, C is constant, and k is the item number of model, A _i, B _iand θ _ifor parameter preset, 0≤i≤k-1, L _n-k+1, L _n-k+2... L _nfor the force value of micro-blog information in front k the second pressure time sequence, ε _n-k+1, ε _n-k+2..., ε _n, ε _n+1be respectively and make E (ε _n-k+1)=0, E (ε _n-k+2)=0 ..., E (ε _n)=0, E (ε _n+1the white noise error item of)=0; X _n-k+1, X _n-k+2... X _nfor micro-blog information in front k the second pressure time sequence force value accumulative and, or be the accumulative of the force value of micro-blog information in front k the second pressure time sequence and and pressure microblogging ratio, or be the pressure microblogging ratio of force value and the number of micro-blog information of micro-blog information in the individual second pressure time sequence of front k.

4. method according to claim 1, is characterized in that, after according to the force value of described time series models determination destination object micro-blog information in next second pressure time sequence, comprising:

According to force value and the corresponding pressure events of micro-blog information in described each second pressure time sequence, build-up pressure changing pattern collection model;

The micro-blog information issued according to described destination object and the force value of micro-blog information, obtain the initial time of goal pressure event corresponding to micro-blog information and described goal pressure event;

According to the type of described goal pressure event, described goal pressure event is mated with described pressure changing pattern collection model, obtains the pressure changing pattern that described goal pressure event is corresponding;

According to the initial time of described goal pressure event, by the pressure changing pattern of described goal pressure event, obtain the modified value of described goal pressure event to described destination object force value of micro-blog information in next second pressure time sequence;

According to the modified value of described goal pressure event to described destination object force value of micro-blog information in next second pressure time sequence, the force value of destination object micro-blog information in next second pressure time sequence that described time series models are determined is revised, obtains the forecast pressure value of described destination object micro-blog information in next second pressure time sequence.

5. method according to claim 4, it is characterized in that, described according to the modified value of described goal pressure event to destination object force value of micro-blog information in next second pressure time sequence, the force value of destination object micro-blog information in next second pressure time sequence that described time series models are determined is revised, comprising:

The pressure changing pattern corresponding to described goal pressure event divides according to event evolves period, obtains the modified value of described pressure changing pattern at different times;

The micro-blog information corresponding according to described goal pressure event, obtains the developing stage of described goal pressure event;

The different times of described pressure changing pattern is mated with the developing stage of described goal pressure event, obtains the modified value matched with the developing stage of described goal pressure event;

According to the modified value that the developing stage of described and described goal pressure event matches, superposition correction is carried out to the force value of destination object micro-blog information in next second pressure time sequence that described time series models are determined.

6., based on a psychological pressure value prediction system for microblogging, it is characterized in that, comprising:

Detection module, detects for the micro-blog information issued destination object, obtains force value and the time of origin of every bar micro-blog information;

First sets up module, for according to the force value of described every bar micro-blog information and time of origin, sets up the first pressure time sequence of described destination object;

Dividing module, for dividing according to Preset Time granularity described first pressure time sequence, obtaining the second pressure time sequence that multiple and described Preset Time granularity is corresponding;

Second sets up module, for obtaining the force value of micro-blog information in described each second pressure time sequence, and according to force value series model Time Created of micro-blog information in described each second pressure time sequence;

Determination module, for the force value of micro-blog information of answering at next second pressure time sequence pair according to described time series models determination destination object.

7. system according to claim 6, is characterized in that, second set up module for:

Gather maximal value, minimum value, the mean value of the force value of micro-blog information in each second pressure time sequence;

According to the accumulative of the maximal value of the force value of micro-blog information in all second pressure time sequences and force value and, set up described time series models;

Or

In all second pressure time sequences, the total number of the minimum value of the force value of micro-blog information, pressure microblogging ratio and microblogging, sets up described time series models;

Or

In all second pressure time sequences the mean value of the force value of micro-blog information, pressure microblogging ratio and force value accumulative and, set up described time series models;

Wherein, the Cumulate Sum of described force value be the force value of micro-blog information in all second pressure time sequences and accumulative and, the total number of described microblogging is the Cumulate Sum of the quantity of the micro-blog information issued in all second time serieses, described pressure microblogging ratio be in each second time series pressure microblogging ratio and Cumulate Sum, in described each second time series, pressure microblogging ratio is the quantity that the force value of micro-blog information in each second time series is greater than micro-blog information in the quantity of the micro-blog information of zero and each second time series.

8. system according to claim 7, is characterized in that, described time series models are:

$L_{n+1}=C+\underset{i=0}{\overset{k-1}{\&Sigma;}}{A}_i{L}_{n-i}+\underset{i=0}{\overset{k-1}{\&Sigma;}}{B}_i{X}_{n-i}+\underset{i=0}{\overset{k-1}{\&Sigma;}}{\mathrm{\&theta;}}_i{\mathrm{\&epsiv;}}_{n-i}+{\mathrm{\&epsiv;}}_{n+1}$

Wherein, L _n+1for the force value of the micro-blog information that destination object is answered at next second pressure time sequence pair, C is constant, L _n-k+1, L _n-k+2... L _nfor the force value of micro-blog information in front k the second pressure time sequence, A _i, B _iand θ _ifor parameter preset, k is the order of model, ε _ifor making E (ε _ithe white noise error item of)=0; X _n-k+1, X _n-k+2... X _nfor micro-blog information in front k the second pressure time sequence force value accumulative and, or, for micro-blog information in front k the second pressure time sequence force value accumulative and and pressure microblogging ratio; Or, be the pressure microblogging ratio of force value and the number of micro-blog information of micro-blog information in front k the second pressure time sequence.

9. system according to claim 6, is characterized in that, this system also comprises:

Correcting module, for according to the force value of micro-blog information in described each second pressure time sequence and corresponding pressure events, build-up pressure changing pattern collection model;

The micro-blog information issued according to described destination object and the force value of micro-blog information, obtain the initial time of goal pressure event corresponding to micro-blog information and described goal pressure event;

According to the type of described goal pressure event, described goal pressure event is mated with described pressure changing pattern collection model, obtains the pressure changing pattern that described goal pressure event is corresponding;

According to the initial time of described goal pressure event, by the pressure changing pattern of described goal pressure event, obtain the modified value of described goal pressure event to described destination object force value of micro-blog information in next second pressure time sequence;

According to the modified value of described goal pressure event to described destination object force value of micro-blog information in next second pressure time sequence, the force value of destination object micro-blog information in next second pressure time sequence that described time series models are determined is revised, obtains the forecast pressure value of described destination object micro-blog information in next second pressure time sequence.

10. system according to claim 9, is characterized in that, described correcting module is used for:

The pressure changing pattern corresponding to described goal pressure event divides according to event evolves period, obtains the modified value of described pressure changing pattern at different times;

The micro-blog information corresponding according to described goal pressure event, obtains the developing stage of described goal pressure event;

The different times of described pressure changing pattern is mated with the developing stage of described goal pressure event, obtains the modified value matched with the developing stage of described goal pressure event;

According to the modified value that the developing stage of described and described goal pressure event matches, superposition correction is carried out to the force value of destination object micro-blog information in next second pressure time sequence that described time series models are determined.