CN104820723A - Online vehicle recommending and sequencing method on basis of preference learning of owners of goods - Google Patents

Abstract

The invention provides an online vehicle recommending and sequencing method on the basis of preference learning of owners of goods. The online vehicle recommending and sequencing method comprises the following steps: firstly establishing a full-connection diagram according to related attribute information of all vehicles; then calculating a probability matrix and an expected matrix of browsed vehicles according to the vehicles which are browsed by the owners of goods and attract the interest of the owners of goods; calculating a matching-degree vector of the vehicles not browsed by the owners of goods and the expected vehicles of the owners of goods, wherein each component of the matching-degree vector represents the matching degree of the corresponding vehicles and the expected vehicles of the owners of goods; and finally, sequencing the vehicles not browsed by the owners of goods according to the matching degree, and repeating the process till the owners of goods select the expected vehicles. The online vehicle recommending and sequencing method provided by the invention has the advantages that the vehicles selected subsequently by users are updated and sequenced by online learning to the preference information of the users, so that the matching degree between the vehicle sequencing result and the user expect is increased, the user experience is improved and the vehicle searching efficiency is improved.

Description

A kind of vehicle based on the study of owner of cargo's Preference recommends sort method online

Technical field

The present invention relates to infotech and automation field, particularly a kind of vehicle based on the study of owner of cargo's Preference recommends sort method online.

Background technology

Logistics platform establishes the information bridge between car owner and the owner of cargo, and car owner can carry out the issue of information of vehicles and searching of goods, and the owner of cargo also can carry out the issue of goods information and vehicle is searched.The owner of cargo carry out vehicle search time, the owner of cargo can consider the impact of the many factors such as safety, driver's credibility, traveling state of vehicle, maintenance situation, even if under the condition met the demands equally, the owner of cargo also has obvious Preference to the selection of vehicle.But present logistics platform scarcely provides query option or only provides several simple query option, inquiry rear vehicle ranking results is difficult to consistent with the expectation of the owner of cargo, and ranking results is unsatisfactory; The Preference also do not selected vehicle according to the owner of cargo at present, carries out the research of other vehicle auto-sequencing aspects in vehicle pond.How to improve vehicle in vehicle pond to sort and matching degree between user's expectation, realize the on-line study of user preference, improving the owner of cargo to the satisfaction of Query Result, is technological difficulties urgently to be resolved hurrily.

Summary of the invention

To the object of the invention is to overcome on existing logistics platform the owner of cargo when searching vehicle, the problem that in the vehicle pond existed, between vehicle sequence and user's expectation, matching degree is not high, propose a kind of vehicle based on the study of owner of cargo's Preference and recommend sort method online, the method is based on the Preference of the owner of cargo, auto-sequencing can be carried out to vehicle in vehicle pond, improve speed and the satisfaction of owner of cargo's enquiring vehicle.

To achieve these goals, the invention provides a kind of vehicle based on the study of owner of cargo's Preference and recommend sort method online, described method comprises: first set up full connection layout according to the correlation attribute information of all vehicles; Then that browsed according to the owner of cargo with interested vehicle calculating probability matrix and viewed vehicle expected matrix; Thus the not viewed vehicle of calculating and the owner of cargo expect the matching degree vector of vehicle, each component of described matching degree vector characterizes the matching degree that corresponding vehicle and the owner of cargo expect vehicle, finally according to matching degree, not viewed vehicle is sorted, repeatedly carry out said process, until the owner of cargo chooses the vehicle of expectation.

In technique scheme, described method specifically comprises the following steps:

Step 1) obtain the correlation attribute information of n chassis to be checked; And set up the full connection layout of vehicle;

Step 2) browse m chassis as the owner of cargo, and when clicking the interested vehicle of the k platform checked in m chassis, calculating probability matrix and the owner of cargo are to the expectation value matrix of a m chassis browsed;

Step 3) calculate not viewed n-m chassis and expect that the matching degree of vehicle is vectorial with the owner of cargo;

Step 4) to step 3) n-m component in the matching degree vector that obtains sort from big to small, and sort to not viewed n-m chassis according to the order of correspondence;

Step 5) if the owner of cargo does not complete the selection of vehicle, make m=m+1, proceed to step 3), until the owner of cargo chooses the vehicle of expectation.

In technique scheme, described step 1) in correlation attribute information comprise: the distance of vehicle and the owner of cargo, vehicle load, tonnage, cumulative volume, current space available volume, length, tenure of use and service time.

In technique scheme, described step 1) in the detailed process setting up the full connection layout of vehicle be:

Calculate the connection weight of vehicle node i to vehicle node j:

w _ij＝exp(-α _i||x _i-x _j||)，i＝1…n,j＝1…n

Wherein i, j represent node numbering, and each node is a chassis, x _i, x _jrepresent the property vector of vehicle node i, j, x _i=(x _i1, x _i2... x _ir), x _j=(x _j1, x _j2..., x _jr), total r attribute in vector, respectively expression: the distance of vehicle and the owner of cargo, vehicle load, tonnage, cumulative volume, current space available volume, length, tenure of use and service time; α _i=(α _i1, α _i2..., α _ir) be given parameter vector, represent the weight of each attribute, α _i1(i=1 ... r) span is (0,1);

According to the connection weight of vehicle node i to vehicle node j, the probability of spreading calculating vehicle node i to vehicle node j is:

$p_{\mathrm{ij}}=\frac{w_{\mathrm{ij}}}{\underset{l=1}{\overset{n}{\mathrm{\Σ}}}{w}_{\mathrm{il}}}.$

In technique scheme, described step 2) detailed process be:

Probability matrix is:

$P=\left[\begin{array}{cc}{P}_{m,m}& P_{m,n-m}\\ P_{n-m,m}& p_{n-m,n-m}\end{array}\right]$

Wherein, P _m,m=[p _ij] _{m × m}by the probability of spreading p of viewed m chassis _ijthe matrix of composition, i=1,2 ..., m, j=1,2 ..., m;

P _{m, n-m}=[p _ij] _{m × (n-m)}by the probability of spreading p between viewed m chassis and not viewed n-m chassis _ijform, i=1,2 ..., m, j=m+1, m+2 ..., n;

P _{n-m, m}matrix P _{m, n-m}transposed matrix;

P _{n-m, n-m}=[p _ij] _{(n-m) × (n-m)}by the probability of spreading p of not viewed n-m chassis _ijthe matrix of composition, i=m+1, m+2 ..., n, j=m+1, m+2 ..., n;

The expectation value matrix of the owner of cargo to a m chassis browsed is:

Y _m＝[y _i] _m×1

Wherein, when i is the sequence number clicking the interested vehicle of k platform checked, y _i=1, otherwise, y _i=0.

In technique scheme, described step 3) the computation process of matching degree vector be:

F _n-m＝(I _n-m,n-m-P _n-m,n-m) ^-1P _n-m,mY _m

Wherein, I _{n-m, n-m}for unit matrix; F _n-mfor the matching degree vector of n-m dimension, each component of described matching degree vector characterizes the matching degree that corresponding vehicle and the owner of cargo expect vehicle.

The invention has the advantages that: by the preference information of on-line study user, for user carries out more new sort to the vehicle of follow-up selection, improve vehicle ranking results and user expect between matching degree, improve Consumer's Experience, improve vehicle search efficiency.

Accompanying drawing explanation

Fig. 1 is the process flow diagram that sort method recommended online by the vehicle based on the study of owner of cargo's Preference of the present invention.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.

Step 1) obtain all vehicle correlation attribute information to be checked; And set up the full connection layout of vehicle;

All vehicles to be checked add up to n, the correlation attribute information of each vehicle comprises: the distance of vehicle and the owner of cargo, vehicle load, tonnage, cumulative volume, current space available volume, length, tenure of use and service time;

Calculate the connection weight of vehicle node i to vehicle node j:

w _ij＝exp(-α _i||x _i-x _j||)，i＝1…n,j＝1…n

Wherein i, j represent node numbering, and each node is a chassis, x _i, x _jrepresent the property vector of vehicle node i, j, x _i=(x _i1, x _i2... x _ir), x _j=(x _j1, x _j2..., x _jr), total r attribute in vector, respectively expression: the distance of vehicle and the owner of cargo, vehicle load, tonnage, cumulative volume, current space available volume, length, tenure of use and service time; α _i=(α _i1, α _i2..., α _ir) be given parameter vector, represent the weight of each attribute, α _i1(i=1 ... r) span is (0,1).

According to the connection weight of vehicle node i to vehicle node j, the probability of spreading calculating vehicle node i to vehicle node j is:

$p_{\mathrm{ij}}=\frac{w_{\mathrm{ij}}}{\underset{l=1}{\overset{n}{\mathrm{\Σ}}}{w}_{\mathrm{il}}}.$

Step 2) browse m chassis as the owner of cargo, and when clicking the interested vehicle of the k platform checked in m chassis, calculating probability matrix and the owner of cargo are to the expectation value matrix of a m chassis browsed;

Probability matrix is:

$P=\left[\begin{array}{cc}{P}_{m,m}& P_{m,n-m}\\ P_{n-m,m}& p_{n-m,n-m}\end{array}\right]$

Wherein, P _m,m=[p _ij] _{m × m}by the probability of spreading p of viewed m chassis _ijthe matrix of composition, i=1,2 ..., m, j=1,2 ..., m;

P _{m, n-m}=[p _ij] _{m × (n-m)}by the probability of spreading p between viewed m chassis and not viewed n-m chassis _ijform, i=1,2 ..., m, j=m+1, m+2 ..., n;

P _{n-m, m}matrix P _{m, n-m}transposed matrix;

P _{n-m, n-m}=[p _ij] _{(n-m) × (n-m)}by the probability of spreading p of not viewed n-m chassis _ijthe matrix of composition, i=m+1, m+2 ..., n, j=m+1, m+2 ..., n;

The expectation value matrix of the owner of cargo to a m chassis browsed is:

Y _m＝[y _i] _m×1

Wherein, when i is the sequence number clicking the interested vehicle of k platform checked, y _i=1, otherwise, y _i=0.

Step 3) calculate not viewed vehicle and the owner of cargo and expect that the matching degree of vehicle is vectorial:

F _n-m＝(I _n-m,n-m-P _n-m,n-m) ^-1P _n-m,mY _m

Wherein, I _{n-m, n-m}for unit matrix; F _n-mfor the matching degree vector of n-m dimension, each component of described matching degree vector characterizes the matching degree that corresponding vehicle and the owner of cargo expect vehicle;

Step 4) to matching degree vector F _n-ma middle n-m component sorts from big to small, and sorts according to the carrying out of order to not viewed n-m chassis of correspondence;

Step 5) if the owner of cargo does not complete the selection of vehicle, make m=m+1, proceed to step 3), until the owner of cargo chooses the vehicle of expectation.

Claims (6)

1. the vehicle based on the study of owner of cargo's Preference recommends a sort method online, and described method comprises: first set up full connection layout according to the correlation attribute information of all vehicles; Then that browsed according to the owner of cargo with interested vehicle calculating probability matrix and viewed vehicle expected matrix; Thus calculate not viewed vehicle and the owner of cargo and expect that the matching degree of vehicle is vectorial, each component of described matching degree vector characterizes the matching degree that corresponding vehicle and the owner of cargo expect vehicle; Finally according to matching degree, not viewed vehicle is sorted, repeatedly carry out said process, until the owner of cargo chooses the vehicle of expectation.

2. the vehicle based on the study of owner of cargo's Preference according to claim 1 recommends sort method online, and described method specifically comprises:

Step 1) obtain the correlation attribute information of n chassis to be checked; And set up the full connection layout of vehicle;

Step 2) browse m chassis as the owner of cargo, and when clicking the interested vehicle of the k platform checked in m chassis, calculating probability matrix and the owner of cargo are to the expectation value matrix of a m chassis browsed;

Step 3) calculate not viewed n-m chassis and expect that the matching degree of vehicle is vectorial with the owner of cargo;

Step 4) to step 3) n-m component in the matching degree vector that obtains sort from big to small, and sort to not viewed n-m chassis according to the order of correspondence;

Step 5) if the owner of cargo does not complete the selection of vehicle, make m=m+1, proceed to step 3), until the owner of cargo chooses the vehicle of expectation.

3. the vehicle based on the study of owner of cargo's Preference according to claim 2 recommends sort method online, it is characterized in that, described step 1) in correlation attribute information comprise: the distance of vehicle and the owner of cargo, vehicle load, tonnage, cumulative volume, current space available volume, length, tenure of use and service time.

4. according to claim 3 based on the owner of cargo's Preference study vehicle recommend sort method online, it is characterized in that, described step 1) in the detailed process setting up the full connection layout of vehicle be:

Calculate the connection weight of vehicle node i to vehicle node j:

w _ij＝exp(-α _i||x _i-x _j||)，i＝1…n,j＝1…n

Wherein i, j represent node numbering, and each node is a chassis, x _i, x _jrepresent the property vector of vehicle node i, j, x _i=(x _i1, x _i2... x _ir), x _j=(x _j1, x _j2..., x _jr), total r attribute in vector, respectively expression: the distance of vehicle and the owner of cargo, vehicle load, tonnage, cumulative volume, current space available volume, length, tenure of use and service time; α _i=(α _i1, α _i2..., α _ir) be given parameter vector, represent the weight of each attribute, α _i1(i=1 ... r) span is (0,1);

According to the connection weight of vehicle node i to vehicle node j, the probability of spreading calculating vehicle node i to vehicle node j is:

$p_{\mathrm{ij}}=\frac{w_{\mathrm{ij}}}{\underset{l=1}{\overset{n}{\mathrm{\Σ}}}{w}_{\mathrm{il}}}.$

5. according to claim 4 based on the owner of cargo's Preference study vehicle recommend sort method online, it is characterized in that, described step 2) detailed process be:

Probability matrix is:

$P=\left[\begin{array}{cc}{P}_{m,m}& P_{m,n-m}\\ P_{n-m,m}& p_{n-m,n-m}\end{array}\right]$

Wherein, P _m,m=[p _ij] _{m × m}by the probability of spreading p of viewed m chassis _ijthe matrix of composition, i=1,2 ..., m, j=1,2 ..., m;

P _{m, n-m}=[p _ij] _{m × (n-m)}by the probability of spreading p between viewed m chassis and not viewed n-m chassis _ijform, i=1,2 ..., m, j=m+1, m+2 ..., n;

P _{n-m, m}matrix P _{m, n-m}transposed matrix;

P _{n-m, n-m}=[p _ij] _{(n-m) × (n-m)}by the probability of spreading p of not viewed n-m chassis _ijthe matrix of composition, i=m+1, m+2 ..., n, j=m+1, m+2 ..., n;

The expectation value matrix of the owner of cargo to a m chassis browsed is:

Y _m＝[y _i] _m×1

Wherein, when i is the sequence number clicking the interested vehicle of k platform checked, y _i=1, otherwise, y _i=0.

6. the vehicle based on the study of owner of cargo's Preference according to claim 5 recommends sort method online, it is characterized in that, described step 3) the computation process of matching degree vector be:

F _n-m＝(I _n-m,n-m-P _n-m,n-m) ^-1P _n-m,mY _m

Wherein, I _{n-m, n-m}for unit matrix; F _n-mfor the matching degree vector of n-m dimension, each component of described matching degree vector characterizes the matching degree that corresponding vehicle and the owner of cargo expect vehicle.