CN105023177A - Intelligent shopping guiding method - Google Patents

Abstract

The invention discloses an intelligent shopping guiding method, and relates to the fields of a big data analytics technology and a data mining technology. The intelligent shopping guiding method comprises the following steps of: a1, data acquisition: acquiring position data of a mobile terminal by developing a map-based fast wifi positioning fingerprint data acquisition algorithm, and simultaneously acquiring detailed map data in a market; a2, data cleaning: cleaning the acquired data, and eliminating the problems of inconsistency, vacancy and noise in the data; a3, data analysis: performing cluster analysis on the cleaned data, extracting a group track model of consumers, and mining a frequent pattern; a4, intelligent guiding: in combination with application scenarios, consumer tracks and the frequent pattern, performing intelligent guiding. According to the invention, by data mining and analysis, shopping preference of the consumers can be explored, accordingly, intelligent and diversified guiding is performed, consumption satisfaction of the consumers is improved, and new growth points are sought for markets and supermarkets in intense competition.

Description

Intelligent shopping guide method

Technical field

The present invention relates to large data analysis technique, data mining technology field; Be specifically related to a kind of intelligent shopping guide method.

Background technology

Along with the fast development of ecommerce, the market share of retail brick (surpassing hereinafter referred to as business) receives severe crush; Moreover, numerous business surpass between homogeneity competition also to make business surpass customer retaining more and more difficult.In order to the challenge of the impact and colleague of tackling electric business, promoting consumption experience, keeping consumer here is the only way that business surpasses.

Consider that the universal people of enabling of smart mobile phone obtain the real time position data of user easily, imply abundant user behavior rule information in these data behind.Therefore, in commercial environments, deep excavation and analysis are carried out to the space time information of strolling shop consumer, us can be made more profoundly to understand shop of the strolling behavioural characteristic of individual consumer and the preference to product and retail shop, and then to Products Show, retail shop is recommended, and retail shop's layout and advertisement pushing etc. have very important significance.

At present, under business environment, mainly concentrate on location navigation aspect for the application of consumer's space time information, such as: consumer is navigated to the anywheres such as the super interior elevator of business, shop, cashier; In addition, the statistical study based on space-time data also can be directly perceived, shows all sidedly and stroll the real time position of shop consumer, residence time, the visiting frequency and the super interior intensity of passenger flow of business, passenger flow moving-wire etc.

Although more than application is based upon on the basis to the initial analysis of consumer's space time information, up to now, we not yet find that correlation technique is analysed in depth the knowledge that these space time informations are contained behind, excavate the shopping preferences of consumer, and carry out intellectuality, diversified shopping guide accordingly.But the key that such technology is widely regarded as " improve consumer spending satisfaction, find new growth point for business surpasses in keen competition ".

Summary of the invention

The object of the invention is for Problems existing in above-mentioned background technology, shop of the strolling space time information of consumer is analysed in depth, find the interest-degree of consumer for product, brand etc., and then realize intelligent shopping guide.

In order to reach above-mentioned technique effect, the present invention takes following technical scheme: a kind of intelligent shopping guide method, comprises the following steps:

A1, data acquisition: by the quick wifi location fingerprint data gathering algorithm of exploitation based on map, obtain mobile terminal locations data, meanwhile, obtain the detailed map data in market;

A2, data cleansing: clean the data collected, eliminate in data the problem of inconsistent, vacancy, the noise existed;

A3, data analysis: carry out cluster analysis to the data after cleaning, extract colony's locus model of consumer, Mining Frequent Patterns;

A4, intelligent shopping guide: connected applications scene, client's track and frequent mode carry out intelligentized shopping guide.

Further technical scheme is: above-mentioned steps a2 data cleansing comprises the following steps:

A21, employing window smothing filtering mode, remove client's floor location error that data acquisition causes;

A22, arest neighbors judge, theoretical according to Jordan curve, judge customer site whether in certain shop, the Euclidean distance of calculating customer site and shop central point simultaneously, and consumer being included into from its nearest shop, the client that removal data acquisition causes enters shop positioning error.

Further technical scheme is: above-mentioned steps a3 data analysis comprises the following steps:

A31, acquisition client stroll shop track;

A32, hot topic/unexpected winner shop judges;

A33, popular trajectory analysis;

A33, stroll shop interest based on user and carry out collaborative filtering recommending.

Further technical scheme is: described hot topic/unexpected winner shop judges to take following methods: obtain each shop

Spread all ID and number of strolling shop consumer, set high and low two threshold values, when the shop number of strolling is greater than height

Threshold value, illustrates that this shop is popular shop; When the shop number of strolling is less than Low threshold, illustrate that this shop is cold

Doors shop.

Further technical scheme is: described popular trajectory analysis takes following methods:

B1, calculate frequent 2 path collection;

B2, carry out Path extension to each frequent 2 respectively, preserve final path.

Further technical scheme is: step b2 is realized by following steps:

B21, extract this head node of frequent 2 and caudal knot point;

B22, travel through frequent 2 path collection, find the child node of head node;

B23, travel through frequent 2 path collection, find the child node of caudal knot point;

Path after b24, acquisition head node and the expansion of caudal knot point; If path does not change, then preserve this path; Otherwise, if path changes, then turn back to step b21.

Further technical scheme is: describedly stroll shop interest based on user and carry out collaborative filtering recommending and take following methods:

C1, obtain shop that every consumer strolls and stroll the shop time accordingly;

C2, utilize similarity sim between following formulae discovery consumer u and consumer v _{(u, v)}:

${\mathrm{sim}}_{(u,v)}=\frac{{\mathrm{\Σ}}_{i\∈S}(R_{u,i}-\stackrel{\‾}{R_u})(R_{v,i}-\stackrel{\‾}{R_v})}{\sqrt{{\mathrm{\Σ}}_{i\∈S}{(R_{u,i}-\stackrel{\‾}{R_u})}^2}\sqrt{{\mathrm{\Σ}}_{i\∈S}{(R_{v,i}-\stackrel{\‾}{R_v})}^2}}$

R in formula _u,i, R _v,ibe respectively consumer u, v shop of strolling time to shop i, be respectively consumer u, v stroll averaging time in shop;

C3, consumer's similarity is utilized to find out k the consumer the most similar to specifying consumer, and institute's shop of strolling set; That from foregoing assemblage, removes appointment consumer strolls shop, as the set of recommendation shop;

C4, calculating specify consumer to stroll shop time P to recommending the prediction in each shop in the set of shop _u,i

$P_{u,i}=\stackrel{\‾}{R_u}+\frac{{\mathrm{\Σ}}_{v\∈U}(R_{v,i}-\stackrel{\‾}{R_v})*S_{u,v}}{{\mathrm{\Σ}}_{v\∈U}\left(\right|S_{u,v}\left|\right)}$

S in formula _u,vfor the similarity of consumer u, v;

C5, to prediction shop of the strolling time sort from high in the end, and choose top n shop recommend specify consumer.

The present invention compared with prior art, there is following beneficial effect:, deep excavation and analysis are carried out to the space time information of strolling shop consumer, us can be made more profoundly to understand shop of the strolling behavioural characteristic of individual consumer and the preference to product and retail shop, and then to Products Show, retail shop is recommended, and retail shop's layout and advertisement pushing etc. have very important significance.Excavate the shopping preferences of consumer, and carry out intellectuality, diversified shopping guide accordingly, improve consumer spending satisfaction, in keen competition, find new growth point for business surpasses.

Accompanying drawing explanation

Fig. 1 is the inventive method schematic flow sheet;

Fig. 2 is trajectory analysis process flow diagram of the present invention.

Embodiment

Below in conjunction with embodiments of the invention, the invention will be further elaborated.

Embodiment:

As shown in Figure 1, the method for intelligent shopping guide mainly comprises the following steps:

(1) data acquisition: by the quick wifi location fingerprint data gathering algorithm of exploitation based on map, obtain mobile terminal locations data, meanwhile, obtains the detailed map data in market.

(2) data cleansing:

May error be there is in the customer site's data collected by location technology.Therefore, we adopt following methods to carry out data cleansing, to eliminate inconsistent and wrong information, for subsequent analysis provides high-quality data.

(1), for consumer at short notice in the situation that different floor jumps, we adopt the mode of window smothing filtering to carry out filtering and revising.

(2), for same consumer beating heart inside and outside same shop jump out existing situation, and we process more flexibly to the shop behavior that enters of consumer, refers to step and enter shop judgement.

(3) data analysis:

(1), enter shop to judge.Theoretical according to Jordan curve, judge customer site whether in certain shop (shape changeable).Calculate the Euclidean distance of customer site and shop central point simultaneously, and consumer is included into from its nearest shop, as the subsidiary conditions entering shop judgement.

Jordan curve is theoretical, and namely to a closed camber line region, if certain point is in this region, the ray so extended from this to any direction all has intersection point with the limit in this region; Otherwise, if certain point is beyond this region, so all there is no intersection point from this along any limit in the ray in part direction and this region.

Euclidean distance, also namely Euclidean distance (Euclidean distance) also claims Euclidean distance, and it is a distance definition usually adopted, and it is the actual distance in m-dimensional space between two points.Euclidean distance in two and three dimensions space is exactly the distance between 2.

The formula of two dimension

d＝sqrt((x1-x2)^2+(y1-y2)^2)

Three-dimensional formula

d＝sqrt((x1-x2)^2+(y1-y2)^2+(z1-z2)^2)

Be generalized to n-dimensional space,

The formula of Euclidean distance

D=sqrt (∑ (xi1-xi2) ^2) i=1,2..n here

Xi1 represents the i-th dimension coordinate of first point, and xi2 represents the i-th dimension coordinate of second point,

It is a point set that n ties up Euclidean space, its each point can be expressed as (x (1), x (2), ... x (n)), wherein x (i) (i=1,2...n) is real number, be called i-th coordinate of x, distance d (x, y) between two point x and y=(y (1), y (2) ... y (n)) is defined as formula above.

(2), " stroll shop " to judge.First, we to consumer stroll shop and divide: for the shop (S) that consumer's (C) residence time (t) is longer, we are denoted as LS (long stay) class shop; Otherwise for passing through on the way but the substantive shop stopped, we are designated SS (short stay) class shop.Thus, we only judge for LS class shop consumer " strolling shop "; We provide time parameter simultaneously, adjust the definition of LS class shop flexibly.

(3), shop track is strolled in extraction.Obtain the LS class shop that consumer strolls within a continuous print time period, and the track formed by these shops.

(4), hot topic/unexpected winner shop.Obtain all ID (as: mobile phone mac address, represents with UeMac) and the numbers of strolling shop consumer in each shop, set high and low two threshold values.When the shop number of strolling is greater than high threshold, illustrate that this shop is popular shop; When the shop number of strolling is less than Low threshold, illustrate that this shop is unexpected winner point shop.

(5), popular track (Hot Routine) analyze, as shown in Figure 2.

B1, calculate frequent 2 path collection FrequentTwo: add up the next shop entered after all consumers go out shop in each popular shops, when the number entering this shop reaches setting threshold value, then think that these two shops are frequent binomial.Preserve the UeMac address of frequent 2 and correspondence.Finally are sorted according to people's fluxion is descending in frequent 2 paths.

B2, carry out Path extension to each frequent 2 respectively, preserve final path.

B21, extract this head node head of frequent 2 and caudal knot point tail.

B22, travel through frequent 2 path collection, find the child node of head node.

Traversal FrequentTwo, judges whether that the caudal knot point of frequent 2 is head node.If had, judge whether the number that this two paths has reaches setting threshold value, if reach setting threshold value, then thinks that this head node of frequent 2 is the child node in path.

B23, travel through frequent 2 path collection, find the child node of caudal knot point;

Traversal FrequentTwo, judges whether that the head node of frequent 2 is tail node.If had, judge whether the number that this two paths has reaches setting threshold value, if reach setting threshold value, then thinks that this caudal knot of frequent 2 point is the child node in path.

Path after b24, acquisition head, the expansion of caudal knot point; If path does not change, then preserve this path; Otherwise, if path changes, then turn back to step b21.

(6) collaborative filtering recommending of shop interest, is strolled based on user.

C1, obtain shop that every consumer strolls and stroll the shop time accordingly.

C2, utilize similarity sim between following formulae discovery consumer u and consumer v _{(u, v)}:

${\mathrm{sim}}_{(u,v)}=\frac{{\mathrm{\Σ}}_{i\∈S}(R_{u,i}-\stackrel{\‾}{R_u})(R_{v,i}-\stackrel{\‾}{R_v})}{\sqrt{{\mathrm{\Σ}}_{i\∈S}{(R_{u,i}-\stackrel{\‾}{R_u})}^2}\sqrt{{\mathrm{\Σ}}_{i\∈S}{(R_{v,i}-\stackrel{\‾}{R_v})}^2}}$

R in formula _u,i, R _v,ibe respectively u, v shop of strolling time to shop i, be respectively u, v stroll averaging time in shop.

C3, consumer's similarity is utilized to find out k the consumer the most similar to specifying consumer, and institute's shop of strolling set; That from foregoing assemblage, removes appointment consumer strolls shop, as the set of recommendation shop.

C4, calculating specify consumer to stroll shop time P to recommending the prediction in each shop in the set of shop _u,i

$P_{u,i}=\stackrel{\‾}{R_u}+\frac{{\mathrm{\Σ}}_{v\∈U}(R_{v,i}-\stackrel{\‾}{R_v})*S_{u,v}}{{\mathrm{\Σ}}_{v\∈U}\left(\right|S_{u,v}\left|\right)}$

S in formula _u,vfor the similarity of u, v.

C5, to prediction shop of the strolling time sort from high in the end, and choose top n shop recommend specify consumer.

(4) intelligent shopping guide:

We first by abstract for " strolling shop " behavior of single consumer C for shape as fc=S1->S2-> ...->St (wherein Si, i ∈ [1, t] for successively stroll LS class shop) form, it mates with known popular track HT by we subsequently, if fc comprises the continuous substring (we use the LS shop of substring notation connected reference) in HT, so access shop follow-up for expectation consumer C is not yet comprise other shops into fc in HT by we, and it is being pushed to consumer.

Simultaneously we are in conjunction with collaborative filtering, carry out Similarity measures and stay time is predicted according to shop that consumer strolls and stay time, and obtain user's residence time the longest N number of shop on this basis.Recommendation results is finally returned to consumer by us.

Be understandable that, the illustrative embodiments that above embodiment is only used to principle of the present invention is described and adopts, but the present invention is not limited thereto.For those skilled in the art, without departing from the spirit and substance in the present invention, can make various modification and improvement, these modification and improvement are also considered as protection scope of the present invention.

Claims (7)

1. an intelligent shopping guide method, is characterized in that, comprises the following steps:

A1, data acquisition: by the quick wifi location fingerprint data gathering algorithm of exploitation based on map, obtain mobile terminal locations data, meanwhile, obtain the detailed map data in market;

A2, data cleansing: clean the data collected, eliminate in data the problem of inconsistent, vacancy, the noise existed;

A3, data analysis: carry out cluster analysis to the data after cleaning, extract colony's locus model of consumer, Mining Frequent Patterns;

A4, intelligent shopping guide: connected applications scene, client's track and frequent mode carry out intelligentized shopping guide.

2. intelligent shopping guide method according to claim 1, is characterized in that, step a2 data cleansing comprises the following steps:

A21, employing window smothing filtering mode, remove client's floor location error that data acquisition causes;

A22, arest neighbors judge, theoretical according to Jordan curve, judge customer site whether in certain shop, the Euclidean distance of calculating customer site and shop central point simultaneously, and consumer being included into from its nearest shop, the client that removal data acquisition causes enters shop positioning error.

3. intelligent shopping guide method according to claim 1 and 2, is characterized in that, step a3 data analysis comprises the following steps:

A31, acquisition client stroll shop track;

A32, hot topic/unexpected winner shop judges;

A33, popular trajectory analysis;

A33, stroll shop interest based on user and carry out collaborative filtering recommending.

4. intelligent shopping guide method according to claim 3, it is characterized in that, described hot topic/unexpected winner shop judges to take following methods: obtain all ID and the numbers of strolling shop consumer in each shop, set high and low two threshold values, when the shop number of strolling is greater than high threshold, illustrate that this shop is popular shop; When the shop number of strolling is less than Low threshold, illustrate that this shop is unexpected winner point shop.

5. intelligent shopping guide method according to claim 3, is characterized in that, described popular trajectory analysis takes following methods:

B1, calculate frequent 2 path collection;

B2, carry out Path extension to each frequent 2 respectively, preserve final path.

6. intelligent shopping guide method according to claim 5, it is characterized in that, step b2 comprises following steps:

B21, extract this head node of frequent 2 and caudal knot point;

B22, travel through frequent 2 path collection, find the child node of head node;

B23, travel through frequent 2 path collection, find the child node of caudal knot point;

Path after b24, acquisition head node and the expansion of caudal knot point; If path does not change, then preserve this path; Otherwise, if path changes, then turn back to step b21.

7. intelligent shopping guide method according to claim 3, is characterized in that, describedly strolls shop interest based on user and carries out collaborative filtering recommending and take following methods:

C1, obtain shop that every consumer strolls and stroll the shop time accordingly;

C2, utilize similarity sim between following formulae discovery consumer u and consumer v _{(u, v)}:

${\mathrm{sim}}_{(u,v)}=\frac{{\mathrm{\Σ}}_{i\∈S}(R_{u,i}-\stackrel{\‾}{R_u})(R_{v,i}-\stackrel{\‾}{R_v})}{\sqrt{{\mathrm{\Σ}}_{i\∈S}{(R_{u,i}-\stackrel{\‾}{R_u})}^2}\sqrt{{\mathrm{\Σ}}_{i\∈S}{(R_{v,i}-\stackrel{\‾}{R_v})}^2}}$

R in formula _u,i, R _v,ibe respectively consumer u, v shop of strolling time to shop i, , be respectively consumer u, v stroll averaging time in shop;

C3, consumer's similarity is utilized to find out k the consumer the most similar to specifying consumer, and institute's shop of strolling set; That from foregoing assemblage, removes appointment consumer strolls shop, as the set of recommendation shop;

C4, calculating specify consumer to stroll shop time P to recommending the prediction in each shop in the set of shop _u,i

$P_{u,i}=\stackrel{\‾}{R_u}+\frac{{\mathrm{\Σ}}_{v\∈U}(R_{v,i}-\stackrel{\‾}{R_v})*S_{u,v}}{{\mathrm{\Σ}}_{v\∈U}\left(\right|S_{u,v}\left|\right)}$

S in formula _u,vfor the similarity of consumer u, v;

C5, to prediction shop of the strolling time sort from high in the end, and choose top n shop recommend specify consumer.