CN103106259A - Mobile webpage content recommending method based on situation - Google Patents

Abstract

The invention relates to a mobile webpage content recommending method capable of providing a user with personalized webpage browsing on a mobile terminal based on a situation. The mobile webpage content recommending method based on the situation comprises steps as below: recording situation information and type of browsed content of mobile webpage browsing of the user at every time, counting the same webpage browsing history records of the browsing history of the user, calculating similarities of the current situation and the situations in a user preference diaphragm, reordering the user preference diaphragm according to the similarities of the situations from high to low, and recommending relevant webpage contents to the user according to the ordering result. The mobile webpage content recommending method based on the situation enables the situation sensing technology to be integrated into an mobile webpage personalized browsing application, and is capable of providing the user with more accurate personalized information, and further improving user experience.

Description

A kind of mobile webpage content recommendation method based on situation

Technical field

A kind of concrete grammar of commending contents in individualized webpage browsing service process that provides for the user according to current situation on mobile terminal device is provided.

Background technology

Along with the development of society and the progress of science and technology, Internet technology is just flourish with unpredictable speed, and steps into gradually huge numbers of families, is deep in the middle of daily life, study and work.The internet relates to various fields and comprises news, scientific research, education, amusement etc. for people provide a large amount of colourful information, and these Internet resources quantity are also presenting the trend of exponential increase simultaneously.Due to the rapid growth of internet size and scale, its content and type also become and become increasingly complex and variation.During user's browsing page, usually can be flooded by the network information of magnanimity, in the process of seeking own interested content, usually will spend more energy and time.Personalized service is a kind of pointed method of service, customer-centric, and by point of interest or the preference of analysis user, the information that provides and recommend to be correlated with for the user.The Extraordinary network browsing can be ignored other information for the individual consumer only for the user pushes its interested content, and then the situation that information is spread unchecked when alleviating user's browsing page, improves user's experience.

In recent years, universal along with mobile intelligent terminal equipment, by mobile terminal device particularly smart mobile phone accessing Internet obtaining information become more and more general.Screen is little, network connects the characteristics such as Bandwidth-Constrained because mobile intelligent terminal equipment has, and therefore the focus that more becomes research is browsed in the personalization of mobile webpage, on the one hand, facilitates the user to read its interested content, reduces the user and diverts one's attention; On the other hand, only load the interested content of user by server, can accelerate the demonstration of web page contents, also can save flow, save expenses of surfing Internet.

The context aware technology can be experienced for the Extraordinary web page browsing technical support is provided.Situation is to describe the information of substance feature, and entity comprises people, thing and various and user or uses mutual object.Context aware is exactly perception entity or user identity, and what Location the current time, does, the process how to do.And these information need to obtain by multiple means usually, and mobile intelligent terminal equipment provides hardware supported for context aware.The large multi-embedding multiple sensors of mobile intelligent terminal now comprises acceleration transducer, GPS etc.Intelligent terminal can the perception user context, and then analyzes active user and individualized feature thereof, comprises customer location, current time, behavior act etc.By user and situation thereof are analyzed, can obtain user personalized information, finally provide personalized service for the user.During the personalization that the context aware technology is dissolved into mobile webpage was browsed, the movability that can take full advantage of on the one hand mobile terminal was obtained multiple user context, can for the user provide customized information more accurately, further improve the user and experience on the other hand.

The webpage personalization is browsed and how to be realized by recommended technology, traditional recommended technology comprises collaborative filtering recommending, content-based recommendation and knowledge type recommendation etc., and modern recommended technology has comprised the recommendation based on context aware, semantic recommendation, cross-domain recommendation etc.At present some occur and moved the method that the webpage personalization is browsed, mostly be by acquisition of information user preferences such as user's registration information, web page browsing record and comments, a plurality of users are analyzed, adopt the method for collaborative filtering to recommend the interested content of user for the user, as patent 200910089587.7.Patent 201110023436.9 has been added position module on this basis, recommends the related news of current location, permanent residence or potential destination for the user, and does not consider that user's historical position information is to the effect of user preference.In the above patent, the first, the user must register, and by a plurality of registered users are analyzed, adopts the method for collaborative filtering to carry out commending contents for the individual consumer; The second, when obtaining user preference, do not consider user's contextual information, comprise time, place and user's behavior etc.Therefore, if webpage does not provide the user to register or the function such as identification, push customized information more accurately for the individual consumer, further improve the user and experience, need abundant user's contextual information analysis user preference, realize content-based recommendation.

Summary of the invention

The technical matters that solves

For fear of the deficiencies in the prior art part, the present invention proposes a kind of content recommendation method of browsing based on the mobile webpage personalization of situation.The method can take full advantage of mobile device and obtain user context, by scenario analysis user preference and behavior, for the user pushes the interested content of different users, improves user's experience that the user reads mobile webpage under different situations.

Technical scheme

A kind of mobile webpage content recommendation method based on situation is characterized in that step is as follows:

Step 1: the type of the contextual information when recording user is browsed mobile webpage at every turn and institute's browsing content, during a web page contents of the every click of user, the contextual information of current web page and the type of click on content are sent in server database and preserve with user network page browsing historical record; Described contextual information is date, when and where;

Described user network page browsing historical record is expressed as:

R_history＝(C ₁,C ₂,…,C _n,Conent)

Wherein, C _kExpression user's different contextual information, k represents the number of situation, Content is the type of the web page contents clicked of user;

Step 2: the same web page browsing history in user's browsing history is counted, represented that with Clicktimes the user clicks respectively the number of times of dissimilar content under different situations, the user preference table of generation is:

R_preference＝(C ₁,C ₂,…,C _n,Conent，Clicktimes)；

Step 3: in the user opens the process of webpage by mobile terminal device, obtain user's current situation CTX by this equipment _cur=(C ₁, C ₂..., C _n), adopt following context aware degree computing formula to calculate the similarity of situation in current situation and user preference table,

$\mathrm{similarity}(x,y)=\frac{{\mathrm{\Σ}}_{k=1}^n{w}_k{\mathrm{\δ}}_k{s}_k(x,y)}{{\mathrm{\Σ}}_{k=1}^n{\mathrm{\δ}}_k}$

Wherein: s _k(x, y) is minute situation C of situation x and y _kSimilarity s _k(x, y); δ _kRepresent corresponding situation C _kThe validity of data, effectively value is 1, invalid value is 0; Divide situation C _kWeight w on the user preference impact _kExpression;

Step 4: from high to low user preference table is reordered according to the result of calculation of context aware degree, again according to click volume to content orderings different under same situation, in user preference table, the web page contents type of first is user's most interested content under current situation; The data that repeat in content type in filter table obtain the sequence of user preference under current situation.Can be the user according to ranking results and recommend the related web page content.

Described contextual information is obtained by terminal device embedded sensors or clock.

The webpage that described user browses need to carry out predefine, use＜div〉＜/div〉web page contents is divided into different semantic chunks, and carry out mark with label.Web page contents of the every click of user can obtain current content affiliated classification, i.e. Content.

Weight w in described step 3 _kCalculation procedure as follows:

Step (1): calculate the interest ratio

Click volume Clicktime (Content with different content _i) ratio that accounts for total click volume Totalclicktime is the interest ratio Content wherein _iRepresent different web page contents types;

Step (2): calculate situation C _kCorresponding Content _iThe variance of click volume

Step (3): with different content at situation C _kUnder the interest ratio multiply by corresponding variance, can try to achieve situation C _kThe interest variance $V_{C_k}=\mathrm{\Σ}{V}_{C_k\left(C{\mathrm{ontent}}_i\right)}\×{\mathrm{Interest}}_{C{\mathrm{ontent}}_i};$

Step (4): different situation C _kShared weight is

Beneficial effect

The invention has the beneficial effects as follows: a kind of content recommendation method that provides individualized webpage to browse according to current situation for the user on mobile terminal device is provided in the present invention.The context aware technology is dissolved in the personalized browse application of mobile webpage, can be provided customized information more accurately for the user, further improve the user and experience.

Description of drawings

Fig. 1: instance user situation acquisition module block diagram of the present invention;

Fig. 2: the treatment scheme sketch of the method for the invention;

Fig. 3: the treatment scheme detail drawing of the method for the invention.

Embodiment

Now in conjunction with the embodiments, the invention will be further described for accompanying drawing:

The technical solution adopted for the present invention to solve the technical problems comprises the following steps:

Step 1: the type of the contextual information when recording user is browsed mobile webpage at every turn and institute's browsing content.Web page contents of the every click of user, the type of the contextual information such as current date, time, place and click on content all will send in server database and preserve.Recommend to lose efficacy if thereby the data that gather too accurately can bring the problem of Sparse to make, therefore in the process of preliminary image data, need to simply process the data that gather, preserve again afterwards.

Step 2: existing user network page browsing record data are processed and analyzed, calculate the user and browse the click volume of dissimilar content under different situations, click volume is higher, and the user interest degree is higher.Generate user preference table, this table can intuitively be expressed the web page contents which kind of type when and where the user like reading.

Step 3: the context aware degree calculates.

When the user opens webpage at every turn, at first obtain the current situation of user by mobile terminal device, this is to like the precondition of content for the user recommends user under current situation.

After obtaining the current situation of user, carry out similarity with the situation in user preference table and calculate.Therefore comprise the multiple minute situations such as when and where due to situation, when calculating the similarity of x and two situations of y, obtain at first respectively minute context aware degree such as the time similarity of x and y and place similarity.Can understand different minute situations by the analysis user historical data and select the factor of influence of different web pages information for the user, be i.e. the different minute weights of situation in similarity computation process.Finally can calculate by context aware degree computing formula the similarity similarity (x, y) of two situations.

Divide in the computation process of context aware degree, often need to carry out extensive to situation, extensive process is namely progressively to gather the loose whole process that is, the rule extensive according to certain, and it is exactly extensive that adjacent, close or similar data are summarized with same data the process that represents.

Step 4: from high to low user preference table is reordered according to the result of calculation of context aware degree, again according to click volume to content orderings different under same situation, like this, in user preference table, the web page contents type of first is the content that the user is most interested under current situation.Simultaneously, filter out the data that repeat in content type in table, what finally obtain is the sequence of user preference under current situation.According to the sequence of user preferences content type under current situation, can be the user and recommend its interested content.

Specific embodiment:

Be recommended as example to realize by smart mobile phone browsing based on the webpage personalization of context aware, in this example, user context is comprised of date, when and where, represent as Fig. 1, date and time can be respectively calendar and clock by mobile terminal obtain, user locations can or initiatively be inputted position information by the user and obtain by the acquisition for mobile terminal gps data.

The treatment scheme of the recommend method of browsing based on the mobile webpage personalization of context aware of the present invention as shown in Figure 2.The specific implementation step is as follows:

Step 201: obtain user context and web page browsing historical data.

The type of the contextual information when recording user is browsed mobile webpage at every turn and institute's browsing content.In this example, situation comprises date, when and where.Date represents respectively that with 1-7 Monday is to Sunday; Time is to represent with 1-8 respectively being divided into 8 time periods in one day, and 1 represents 0:00-3:00, and 2 represent 3:00-6:00, by that analogy; The place can be inputted to obtain by the user, and when the user opened webpage, the point selection dialog box, selected current site by the user ejectedly.The web page contents type adopts and shifts to an earlier date predefined method, with the division of teaching contents in webpage is＜physical culture amusement, politics, military affairs〉four classes, and carry out mark with label.Web page contents of the every click of user, the label record with current date, time, place and institute's browsing content gets off, and is kept in database with following form:

R_history＝(Date,Time,Location，Conent) (1)

Wherein, Date represents the date, and Time represents the time, and Location represents the place, and Content represents the content type of user's webpage clicking.

Step 202: dynamically generate user preference table.

Before each user opens webpage, all can at first generate user preference table according to user's browsing histories data of recording before, so user preference table is that not timing dynamically updates.User's browsing histories data are added up, and click volume Clicktimes represents that the user clicks the number of times of dissimilar content under identical situation, and the number of clicks bright user that more speaks more is interested in this content, finally can obtain user preference table, is expressed as:

R_preference＝(Date，Time,Location，Conent，Clicktimes) (2)

Step 203: the context aware degree calculates.

Calculate the current situation CTX of user _curSimilarity with situation in user preference table.Situation in user preference table is used CTX _i(i=1,2 ..., n) expression, n represents the number that records of user preference table.Calculate respectively the similarity of situation in current situation and preference table, use Sim _iExpression has:

${\mathrm{Sim}}_i=\mathrm{similarity}({\mathrm{CTX}}_{\mathrm{cur}},{\mathrm{CTX}}_i)=\frac{{\mathrm{\Σ}}_{k=1}^m{w}_k{\mathrm{\δ}}_k{s}_k({\mathrm{CTX}}_{\mathrm{cur}},{\mathrm{CTX}}_i)}{{\mathrm{\Σ}}_{k=1}^m{\mathrm{\δ}}_k}---\left(3\right)$

Wherein, m represents the number of types of situation, in this example, date, three situations of when and where are arranged, so the m value is 3; δ _kThe validity that represents the data of corresponding different situations, because the data of three situations that gather are all effective, δ _kValue all can be made as 1; w _kRepresent the shared weights of different situations.Final context aware degree computing formula is:

${\mathrm{Sim}}_i=\frac{{\mathrm{\Σ}}_{k=1}^3{w}_k{s}_k({\mathrm{CTX}}_{\mathrm{cur}},{\mathrm{CTX}}_i)}{3}---\left(4\right)$

Step 204: interest sequence under the current situation of user;

According to the context aware degree Sim that calculates _i(i=1,2, n) from high to low user preference table is sorted, by the click volume Clicktimes in preference table, same situation different content type is carried out partial ordering, like this, in newly-generated user preference table, the web page contents type of first is the content that the user is most interested under current situation.Simultaneously, filter out the data that repeat in content type in table, what finally obtain is the sequence of user preference under current situation.According to the sequence of user preferences content type under current situation, can be the user and recommend its interested one or one group of content.

Fig. 3 is the treatment scheme detail drawing of the method for the invention,

Wherein, step 203 further comprises step 301,302 and 303.

Step 301: obtain the current situation of user.In the user opens the process of webpage by mobile terminal device, need to obtain user's current situation by this equipment, the situation obtain manner obtains the sight method with step 201.Current situation CTX _curCan be expressed as:

CTX _cur＝(Date,Time,Location) (5)

Step 302: at first distinguish calculation date, three corresponding similarity s of situation of when and where _k(CTX _cur, CTX _i), k=3 wherein.Need to apply to the extensive of situation in computation process.In user's browsing histories data, the date represents respectively Monday to Sunday by 1 to 7, and Mon-Fri can extensively be working day, and Saturday and Sunday can extensively be weekend, and this is that one-level is extensive; On this basis, can be " arbitrary day " with all data are extensive carrying out extensive.In like manner, carry out that one-level is extensive is divided into morning, the morning, afternoon and evening the time, if once carrying out extensively, can be divided into day and night.The place can be divided into amusement and study or indoor and outdoors according to different attribute.

In calculating the process of similarity, at first raw data is compared, identical similarity is 1, difference is carried out one-level extensive, compares again, until final two data after extensive are identical.Often carry out once extensively, similarity will be reduced by half on original basis, if for example carry out one-level extensive after two data identical, the similarity of raw data is 0.5, if the extensive rear data of secondary are identical, the similarity of raw data is 0.25.

User's browsing histories data are carried out statistical study, determine that different situations are for the weight of the impact of user preference.

Take the weight calculation of date Date as example, at first calculate the interest ratio, i.e. the click volume Clicktim (eConten of different contents _iT) account for the ratio of total click volume Total clicktime, content type comprises＜physical culture, amusement, politics, military affairs〉four classes, need to calculate respectively.

${\mathrm{Interest}}_{C{\mathrm{ontent}}_i}=\frac{\mathrm{Clicktime}\left({\mathrm{Content}}_i\right)}{\mathrm{Totalclicktime}}---\left(6\right)$

I is made as 1 to 4, represents successively physical culture, amusement, political and military four class contents.

Calculate the corresponding Content of situation Date _iThe variance of click volume.Date comprises 7 values, the Content that different value is corresponding _iClick volume be respectively X ₁, X ₂..., X ₇, variance can be tried to achieve by following formula:

$V_{\mathrm{Date}\left({\mathrm{Content}}_i\right)={\mathrm{\Σ}}_{p=1}^7{(X_p-\stackrel{\‾}{X})}^2---\left(7\right)}$

Wherein

Be user's click on content Content every day _iMean value.

Interest ratio with different content under situation Date multiply by corresponding variance, can try to achieve the interest variance V of situation Date _Date:

$V_{\mathrm{Date}}={\mathrm{\Σ}}_{i=1}^4{V}_{\mathrm{Date}\left({\mathrm{Content}}_i\right)\×{\mathrm{Interest}}_{{\mathrm{Content}}_i}---\left(8\right)}$

In like manner can try to achieve respectively the interest variance V of time situation Time and place situation Location _TimeAnd V _Location

Like this, the shared weight of date situation Date can be expressed as:

$w_{\mathrm{Date}}=\frac{V_{\mathrm{Date}}}{V_{\mathrm{Date}}+V_{\mathrm{Time}}+V_{\mathrm{Location}}}---\left(9\right)$

In like manner can try to achieve time situation weight w _TimeWith place situation weight w _Location

Step 303: each context aware degree and the weight thereof that obtain in step 302 are brought in context aware degree computing formula, can proper front situation and user preference data in the similarity of arbitrary situation.

Claims (4)

1. mobile webpage content recommendation method based on situation is characterized in that step is as follows:

Step 1: the type of the contextual information when recording user is browsed mobile webpage at every turn and institute's browsing content, during a web page contents of the every click of user, the contextual information of current web page and the type of click on content are sent in server database and preserve with user network page browsing historical record; Described contextual information is date, when and where;

Described user network page browsing historical record is expressed as:

R_history＝(C ₁,C ₂,…,C _n,Conent)

Wherein, C _kExpression user's different contextual information, k represents the number of situation, Content is the type of the web page contents clicked of user;

Step 2: the same web page browsing history in user's browsing history is counted, represented that with Clicktimes the user clicks respectively the number of times of dissimilar content under different situations, the user preference table of generation is:

R_preference＝(C ₁,C ₂,…,C _n,Conent,Clicktimes)；

Step 3: in the user opens the process of webpage by mobile terminal device, obtain user's current situation CTX by this equipment _cur=(C ₁, C ₂..., C _n), adopt following context aware degree computing formula to calculate the similarity of situation in current situation and user preference table,

$\mathrm{similarity}(x,y)=\frac{{\mathrm{\Σ}}_{k=1}^n{w}_k{\mathrm{\δ}}_k{s}_k(x,y)}{{\mathrm{\Σ}}_{k=1}^n{\mathrm{\δ}}_k}$

Wherein: s _k(x, y) is minute situation C of situation x and y _kSimilarity s _k(x, y); δ _kRepresent corresponding situation C _kThe validity of data, effectively value is 1, invalid value is 0; Divide situation C _kWeight w on the user preference impact _kExpression;

Step 4: from high to low user preference table is reordered according to the result of calculation of context aware degree, again according to click volume to content orderings different under same situation, in user preference table, the web page contents type of first is user's most interested content under current situation; The data that repeat in content type in filter table obtain the sequence of user preference under current situation.Can be the user according to ranking results and recommend the related web page content.

2. according to claim 1 based on the mobile webpage content recommendation method of situation, it is characterized in that: described contextual information is obtained by terminal device embedded sensors or clock.

3. according to claim 1 based on the mobile webpage content recommendation method of situation, it is characterized in that: the webpage that described user browses need to carry out predefine, use＜div〉＜/div〉web page contents is divided into different semantic chunks, and carry out mark with label.Web page contents of the every click of user can obtain current content affiliated classification, i.e. Content.

4. according to claim 1 based on the mobile webpage content recommendation method of situation, it is characterized in that: weight w in described step 3 _kCalculation procedure as follows:

Step (1): calculate the interest ratio

Click volume Clicktime (Content with different content _i) ratio that accounts for total click volume Total clicktime is the interest ratio

Content wherein _iRepresent different web page contents types;

Step (2): calculate situation C _kCorresponding Content _iThe variance of click volume

Step (3): with different content at situation C _kUnder the interest ratio multiply by corresponding variance, can try to achieve situation C _kThe interest variance $V_{C_k}=\mathrm{\Σ}{V}_{C_k\left(C{\mathrm{ontent}}_i\right)}\×{\mathrm{Interest}}_{C{\mathrm{ontent}}_i};$

Step (4): different situation C _kShared weight is

Images