CN113063431A - Intelligent recommendation method for sharing bicycle riding route - Google Patents

Abstract

The invention discloses an intelligent recommendation method for sharing a bicycle riding route, which comprises the following steps: collecting historical riding data of a plurality of shared bicycle users, and establishing an original data set; generating an optional path set according to historical riding data; constructing a path utility function; determining the probability of each path in the selectable path set being selected; and constructing a correction matrix according to the user history evaluation, the safety coefficient and the riding continuity degree, wherein the correction matrix is used for correcting the probability value and determining the final probability of each path being selected. The invention can push a route which is relatively satisfied with the user and has higher safety factor for the shared bicycle user, and improves the riding experience of the user so as to achieve the aim of smooth travel of the shared bicycle.

Description

Intelligent recommendation method for sharing bicycle riding route

Technical Field

The invention relates to the field of intelligent transportation, in particular to an intelligent recommendation method for sharing a single-vehicle riding route.

Background

The shared bicycle is a time-sharing rental mode with public bicycles as a main body, and the main content of the shared bicycle is to provide services for placing a large number of public bicycles and a small number of electric vehicles in areas with people flowing in surge, such as campuses, subway stations, bus stations, residential areas, commercial areas, public service areas and the like. The shared bicycle is taken as a new trip mode, is favored by more and more people due to the characteristics of convenient, rapid and low-cost public transportation and the advantages of green, energy conservation and health, is vividly called as the last jigsaw in the transportation industry, and is the best trip selection in short distance of many people. Meanwhile, the problem of the last kilometer of the bus traffic can be effectively solved, the important function of relieving urban traffic jam is achieved, and the overall service level of the urban traffic is improved.

However, while bringing convenience to travel of the majority of citizens and tourists, there are some limitations, for example, it is difficult for a user to plan a route when riding on a completely unfamiliar road, and even an optimal route does not need to be provided, and unreasonable route selection may cause delay of the user time and reduction of riding satisfaction. In terms of considerations, navigation on the market today relies heavily on a single variable-ride time or ride distance-for recommendations of a user's ride route. Although some researches are conducted at home and abroad to discuss related problems, the consideration of some road environment factors such as the number of traffic lights, road gradient and the like on the route selection of a rider is increased, the influence of factors influencing the riding safety of the user such as the type of a non-motor vehicle lane, the parking section in the road, the road grade and the like on the route selection of the rider is considered less in relation to the actual traffic condition in China. In the research process, a researcher usually selects a traditional multi-item location model, but the shared bicycle is used for solving the problem of short trip in a city, namely, the trip range of a user during riding is not large, the phenomenon of repeated road sections of an actual path and an optional path set exists, and the traditional multi-item location model cannot recommend a more accurate path for the user due to lack of consideration of the path overlapping condition.

Disclosure of Invention

The invention aims to overcome the defects in the prior art, provides an intelligent recommendation method for sharing a bicycle riding route, aims to push a route which enables users to be relatively satisfied and has a high safety factor for sharing bicycle users, and improves the riding experience of the users so as to achieve the aim of smooth traveling of the sharing bicycle.

In order to achieve the purpose, the invention adopts the following technical scheme:

the invention discloses an intelligent recommendation method for sharing a bicycle riding route, which is characterized by comprising the following steps:

step 1, collecting historical riding data of a plurality of shared bicycle users, and establishing an original data set; the data types in the raw data set include a non-motorway type X₁Parking section X in road₂Traffic lights quantity X₃And riding distance X₄Road grade X₅And road surface gradient X₆；

Step 2, generating an optional path set C with k optional paths according to the historical riding data of the user n_nAnd calculating the user n from the alternative path set C by using the formula (1)_nThe repetition coefficient PS of the selected path i_ni：

In the formula (1), ζ_iIs the set of all road segments in the path i; l_aIs the length of segment a in path i; l is_iIs the length of path i; if the road section a is in the optional path set C_nLet σ be on the path j not selected by the user n_ajIf not, let σ_aj＝0；

Is an alternative path set C_nLength of shortest path on; l is_jIs an alternative path set C_nThe length of the medium path j;

step 3, according to the historical riding data of the user n, assuming that the user n selects a path according to the principle of path utility maximization, constructing a utility function U of the path i by using the formula (2)_ni：

U_ni＝V_ni+ε_ni (2)

In the formula (2), V_niSelecting a path utility value after the riding path i for the user n; epsilon_niSelecting a random utility error of a path i for a user n, and obeying mutually independent Gumbel distribution; and comprises the following components:

in the formula (3), beta₀Is a constant term, X_mData representing an mth data type in the original data set; beta is a_mRepresenting the influence coefficient of the mth data type on the user riding route selection; beta is a_PSRepresenting the influence coefficient of the path repetition coefficient on the user riding route selection;

step 4, obtaining the user n selection selectable path set C according to the formula (4)_nProbability value P (i | C) of medium path i_n) Thus, the alternative path set C can be determined_nThe probability of each path in the set being selected;

step 5, constructing a correction matrix according to the user history evaluation, the safety coefficient and the riding continuity degree, and using the correction matrix to correct the probability value so as to determine the selectable path set C_nThe final probability of each path in the set being selected;

step 5.1, obtaining the optional path set C of other users_nThe evaluation data of each path or road section in the system is normalized after being averaged to obtain an optional path set C_nAverage rating score of each route or section in (1):

step 5.2, calculating a safety factor S_ni：

Constructing a danger rate function h (t | X) sharing the riding time t of the riding path i of the bicycle user n by using the formula (5):

h(t|X)＝h₀(t)g(X) (5)

in the formula (5), h₀(t) reference Risk of riding time t, g(X) represents the effect of factor X on the risk and has:

g(X)＝exp(β′₁X₁+β′₂X₂+β′₃X₃) (6)

in the formula (6),. beta.'₁-β′₃Alternative path sets C respectively representing users n_nType X of non-motor vehicle lane in each path₁Parking section X in road₂And the number of traffic lights X₃Influence coefficient on riding hazard rate;

obtaining a safety factor S of the user n selection path i after the riding time t by using the formula (7)_niThereby obtaining an alternative path set C_nSafety factor of all paths in (1):

step 5.3, calculating the riding continuity Con:

obtaining the optional path set C by using the formula (8)_nThe riding continuity Con of each path:

logCon＝β″₀+β″₃X₃+β″₅X₅+β″₆X₆+ε (8)

in formula (8), β ″)₀Is a constant term; beta ″)₃、β″₅And beta ″)₆Alternative path sets C respectively representing users n_nNumber of traffic lights in each path X₃Road grade X₅And road surface gradient X₆Influence coefficient on the user riding continuity Con; epsilon is a random utility error of the riding continuity Con;

step 5.4, constructing a correction matrix:

the average evaluation score, the safety factor and the riding continuity degree of each path and the probability of each path being selected form a correction matrix

r_1kRepresenting the probability of the k-th path being selected; r is_2kTo representThe user evaluation score of the kth path; r is_3kRepresenting the safety factor of the k path; r is_4kRepresenting the riding continuity degree of the k-th path;

step 5.5, calculating the final probability of each path being selected;

determining alternate path set C using equation (9)_nSelecting the final probability of each path, and recommending the probability maximum value as the optimal path to the user;

[p₁ p₂ … p_k]＝[ω₁ ω₂ ω₃ ω₄]·R (9)

in the formula (9), p_kA final probability of being selected for the kth path; omega₁The probability of each path being selected determined before correction is weighted; omega₂Evaluating the occupied weight for the user; omega₃The safety coefficient is the weight; omega₄Weight of riding continuity, and omega₁+ω₂+ω₃+ω₄＝1。

Compared with the prior art, the invention has the beneficial effects that:

1. the influence of seven significant factors, namely riding distance, the number of traffic lights, types of non-motor vehicles, road grades, parking sections in roads, road surface gradients and path repetition coefficients, on the riding path selection of the shared bicycle user is considered, the riding efficiency, riding safety and riding continuity of the user are considered, and finally a route with the highest pushing probability is pushed for the user to navigate and assist the user in riding, so that the riding experience of the user is improved, the traveling efficiency and riding safety are improved, the traveling mode of the shared bicycle is facilitated, and the shared bicycle can be accepted, used and loved by the public, and the riding method responds to the call of the low-carbon life of the whole population.

2. The invention increases the consideration of more users on the safety of the users by combining the actual conditions of domestic roads, such as the influence of other vehicles on riders in a mixed running state. According to the utility maximization theory, a path utility function is constructed by using a Logit model (PSL) considering the path length, so that a path selection model considering the path overlapping factor is obtained. The model can predict the probability of each route set provided by the user selection system, the attention degree of different users to the user history evaluation, the riding safety factor and the riding continuity degree is considered, a correction matrix is constructed to obtain the final probability of selecting each route set by the user, and finally, the route with the highest probability is pushed for the user to navigate to assist the user in riding. Due to wide and comprehensive consideration factors, the accuracy of the recommended path obtained by the method is higher, and the constructed correction matrix enables the final recommended path to have the characteristic of individuation and enables users to be satisfied better, so that the riding experience of the users is improved.

Drawings

FIG. 1 is a schematic flow chart of an intelligent recommendation method of the present invention;

FIG. 2 is a flow chart illustrating the process of determining the final probability of each path being selected according to the present invention.

Detailed Description

In this embodiment, as shown in fig. 1, an intelligent recommendation method for shared bicycle riding routes includes the following steps:

step 1, collecting historical riding data of a plurality of shared bicycle users, and establishing an original data set; the data types in the raw data set include a non-motorway type X₁Parking section X in road₂Traffic lights quantity X₃And riding distance X₄Road grade X₅And road surface gradient X₆(ii) a Because the explained variables belong to the actual road conditions encountered during riding and have little correlation with personal factors and social factors of riders, the problem of deviation of omitted variables can be avoided.

Wherein the non-motor vehicle lane type X₁The system comprises four types of vehicles and vehicles, namely a vehicle-non-mixed vehicle, a marked line non-motor vehicle lane, a fence isolation non-motor vehicle lane and a greening isolation non-motor vehicle lane, wherein the difference of the types can influence the subjective safety of users. Wherein the non-motor vehicle mixed running means that the non-motor vehicle running in the motor vehicle lane is not divided into non-motor vehicle lanes, but is mostly present on the road surface with narrower road width, and the marked lines, fences and greening are three different isolation devices which are adopted in China and separate the non-motor vehicle lanes from the motor vehicle lanesApplying; in-road parking section X₂The motor vehicle parking facility is a road section with motor vehicle parking facilities, and the motor vehicles in the area generally provided with the motor vehicle parking facilities need to pass through a non-motor lane for parking, sometimes the condition that the vehicles are parked in the non-motor lane occurs, so that dangerous factors can be added to riding; number of traffic lights X₃The number of intersections which the user needs to pass through when riding is strongly positively correlated, and the value indirectly reflects the continuity of the user when riding; riding distance X₄The actual riding path length from the starting point to the returning point is taken as the user; road grade X₅The method comprises the following steps of considering roads allowing non-motor vehicles to pass in a city and considering the characteristic that shared single-vehicle parking points are mostly located outside residential areas or industrial areas, classifying the roads into three types of main roads, secondary main roads and branch roads, wherein the main roads are main roads connecting all subareas of the city, mainly take traffic functions as the main roads, have more and dense motor vehicles, and have the average driving speed of 50 km/h; the secondary trunk road and the main trunk road are combined to form a trunk network, the trunk network has two functions of distributed traffic and service, and the average driving speed of the motor vehicle is 40 km/h; the branch road is connected with the secondary main road and the internal roads of residential areas, industrial areas and the like, is used for solving the problem of roads in local areas, and usually has a service function as a main part, although the number of motor vehicles is less, the number of pedestrians is more, and the average number of motor vehicle traveling elements is 30 km/h; road surface gradient X₆Reflecting the riding difficulty of the road surface, most of users using the shared bicycle are not professional riders, and tend to select a path with a smooth road surface and easiness in riding. The user riding history data X₁-X₆Available from shared bicycle companies, third party apps, or a number of practical investigations.

Step 2, generating an optional path set C with k optional paths according to the historical riding data of the user n_n；

Optional Path set C_nA path in (1) generally refers to a path that a user may select without unreasonable detour behavior. The accuracy of the generated alternative path set is directly related to the calculation of the subsequent path repetition coefficient and the prediction of the user riding path selection probability. The Dijkstra algorithm, A, can be generally adopted for constructing the path selection set^*And path generation algorithms such as an algorithm and a branch and bound algorithm. It should be noted that the method can select a set of pathsThe determination of (a) needs to be modified based on the algorithms to meet the requirement of generating k optional paths. If the shortest path is found, a certain edge may be deleted, and after the new graph is obtained, the shortest path in the new graph, that is, the secondary short path in the original graph, is continuously found. The repeated operation can obtain a series of secondary short paths to form the optional path set of the user n. The route selection range is defaulted to 4km, but if the shortest route in the selectable route set is still larger than 4km, the route selection range can be reasonably expanded, and the total number of the routes in the selectable route set is not more than 7.

Step 3, calculating the optional path set C of the user n_nThe repetition coefficient PS of the selected path i_ni：

The path repetition coefficient is a variable for measuring the repetition condition between the actual selected path and the selectable path set of the user, and is expressed as the behavior of the user facing the audience or the evasion of a large number of riders, and the method utilizes the formula (1) to calculate:

in the formula (1), ζ_iIs the set of all road segments in the path i; l_aIs the length of segment a in path i; l is_iIs the length of path i; if the road section a is in the optional path set C_nLet σ be on the path j not selected by the user n_ajIf not, let σ_aj＝0；

Is an alternative path set C_nLength of shortest path on; l is_jIs an alternative path set C_nThe length of the medium path j;

step 4, according to the historical riding data of the user n, assuming that the user n selects a path according to the principle of path utility maximization, constructing a utility function U of the path i by using the formula (2)_ni：

U_ni＝V_ni+ε_ni (2)

In the formula (2), V_niSelecting a path utility value after the riding path i for the user n; epsilon_niSelecting random utility error of path i for user n, and obeying Gumbel distribution independent from each other, comparing with normal distribution or uniform distribution, adding epsilon obeying Gumbel distribution_niCan make path utility function U_niThe fitting of (2) is closer to the real situation, is favorable to improving the prediction precision of the later stage, and has:

in the formula (3), beta₀Is a constant term, X_mData representing an mth data type in the original data set; beta is a_mRepresenting the influence coefficient of the mth data type on the user riding route selection; beta is a_PSRepresenting the influence coefficient of the path repetition coefficient on the user riding route selection; log (PS)_ni) Indicating path overlap, i.e. path i and alternative path set C_nThe degree of overlap of other paths;

according to the raw data collected in the step 1, dividing the raw data set into different data sets according to the number k of the selectable path sets by using Excel. And then utilizing stata statistical analysis software to obtain 7 interpretation variables of the riding distance, the number of traffic lights, the types of non-motor vehicles, the road grade, the parking sections in the road, the road surface gradient and the path repetition coefficient, wherein the P values are all less than the significance level 0.05, namely the variables have a correlation with the selection of the riding route of the user with the interpreted variables. The parameter value beta can also be estimated by a maximum likelihood function estimation method₀,β₁-β₆,β_PS. Taking the first path as a basic selection item, k functions, namely path utility functions of k paths in the path selection set, can be constructed.

Step 5, obtaining the user n selection selectable path set C according to the formula (4)_nProbability value P (i | C) of medium path i_n) Thus, the alternative path set C can be determined_nThe probability of each path in the set being selected;

step 6, constructing a correction matrix according to the user history evaluation, the safety coefficient and the riding continuity degree, and using the correction matrix to correct the probability value so as to determine the selectable path set C_nThe final probability of each path being selected, the steps of which are shown in fig. 2;

step 6.1, obtaining the optional path set C of other users_nAnd averaging the evaluation data of each path or section and then normalizing. The evaluation scores of other paths in the path set are replaced by the evaluation scores of the road sections related to the paths, if the path lacks the evaluation scores of the related road sections, the path is fully scored, and the selectable path set C is obtained_nAverage rating score for each path in (1):

step 6.2, calculating the safety factor S_ni：

Let t be a continuous time variable representing the riding time of the user n selecting the path i, selecting the non-motor lane type X₁Parking section X in road₂And the number of traffic lights X₃As covariates, a hazard rate function h (t | X) sharing the riding time t of the bicycle user n riding path i is constructed using equation (5):

h(t|X)＝h₀(t)g(X) (5)

in the formula (5), h₀(t) is the reference risk of riding time t, g (X) represents the influence of factor X on the risk, namely neglecting X₁-X₃The risk when g (x) is 1; reference Cox proportional hazards model construction g (x) satisfies formula (6):

g(X)＝exp(β′₁X₁+β′₂X₂+β′₃X₃) (6)

in the formula (6),. beta.'₁-β′₃Alternative path sets C respectively representing users n_nType X of non-motor vehicle lane in each path₁Parking section X in road₂And the number of traffic lights X₃Influence coefficient on riding hazard rate;

the safety factor is in a descending trend along with the increase of the riding time, and the safety factor S of the user n after the riding time t is obtained by utilizing the relational construction formula (7)_niThereby obtaining the alternative pathDiameter set C_nSafety factor of all paths in (1):

step 6.3, calculating the riding continuity Con:

obtaining the optional path set C by using the formula (8)_nThe riding continuity Con of each path:

logCon＝β″₀+β″₃X₃+β″₅X₅+β″₆X₆+ε (8)

in formula (8), β ″)₀Is a constant term; beta ″)₃、β″₅And beta ″)₆Alternative path sets C respectively representing users n_nNumber of traffic lights in each path X₃Road grade X₅And road surface gradient X₆Influence coefficient on the user riding continuity Con; epsilon is a random utility error of the riding continuity Con;

step 6.4, constructing a correction matrix:

the average evaluation score, the safety factor and the riding continuity degree of each path and the probability of each path being selected form a correction matrix

r_1kRepresenting the probability of the k-th path being selected; r is_2kA user rating score representing a kth path; r is_3kRepresenting the safety factor of the k path; r is_4kRepresenting the riding continuity degree of the k-th path;

6.5, calculating the final probability of each path being selected;

determining alternate path set C using equation (9)_nSelecting the final probability of each path, and recommending the probability maximum value as the optimal path to the user;

[p₁ p₂ … p_k]＝[ω₁ ω₂ ω₃ ω₄]·R (9)

in the formula (9)，p_kA final probability of being selected for the kth path; omega₁The probability of each path being selected determined before correction is weighted; omega₂Evaluating the occupied weight for the user; omega₃The safety coefficient is the weight; omega₄Weight of riding continuity, where ω₁+ω₂+ω₃+ω₄＝1。

Step 7, optimizing the method;

it is contemplated that users who are partially familiar with roads may not require path recommendations and some special cases such as the user temporarily changing destinations, the user interrupting the ride, etc. After the method is implemented, the evaluation of the user on the pushing path or the related road section of the method needs to be collected, so that the construction of the later-period correction matrix is facilitated. And comparing the actual selected path of the user with the recommended path of the system, and estimating the accuracy of the method. And meanwhile, the actual riding data of the user is named and filed, and a training set is added, so that the use and analysis of the future optimization method are facilitated.

Claims (1)

1. An intelligent recommendation method for sharing bicycle riding routes is characterized by comprising the following steps:

step 1, collecting historical riding data of a plurality of shared bicycle users, and establishing an original data set; the data types in the raw data set include a non-motorway type X₁Parking section X in road₂Traffic lights quantity X₃And riding distance X₄Road grade X₅And road surface gradient X₆；

Step 2, generating an optional path set C with k optional paths according to the historical riding data of the user n_nAnd calculating the user n from the alternative path set C by using the formula (1)_nThe repetition coefficient PS of the selected path i_ni：

In the formula (1), ζ_iIs the set of all road segments in the path i; l_aIs path i middle wayThe length of segment a; l is_iIs the length of path i; if the road section a is in the optional path set C_nLet σ be on the path j not selected by the user n_ajIf not, let σ_aj＝0；

Is an alternative path set C_nLength of shortest path on; l is_jIs an alternative path set C_nThe length of the medium path j;

step 3, according to the historical riding data of the user n, assuming that the user n selects a path according to the principle of path utility maximization, constructing a utility function U of the path i by using the formula (2)_ni：

U_ni＝V_ni+ε_ni (2)

In the formula (2), V_niSelecting a path utility value after the riding path i for the user n; epsilon_niSelecting a random utility error of a path i for a user n, and obeying mutually independent Gumbel distribution; and comprises the following components:

in the formula (3), beta₀Is a constant term, X_mData representing an mth data type in the original data set; beta is a_mRepresenting the influence coefficient of the mth data type on the user riding route selection; beta is a_PSRepresenting the influence coefficient of the path repetition coefficient on the user riding route selection;

step 4, obtaining the user n selection selectable path set C according to the formula (4)_nProbability value P (i | C) of medium path i_n) Thus, the alternative path set C can be determined_nThe probability of each path in the set being selected;

step 5, constructing a correction moment according to the historical evaluation of the user, the safety coefficient and the riding continuity degreeArray for modifying probability values to determine alternative path sets C_nThe final probability of each path in the set being selected;

step 5.1, obtaining the optional path set C of other users_nThe evaluation data of each path or road section in the system is normalized after being averaged to obtain an optional path set C_nAverage rating score of each route or section in (1):

step 5.2, calculating a safety factor S_ni：

Constructing a danger rate function h (t | X) sharing the riding time t of the riding path i of the bicycle user n by using the formula (5):

h(t|X)＝h₀(t)g(X) (5)

in the formula (5), h₀(t) is the reference risk for riding time t, g (X) represents the effect of factor X on the risk and has:

g(X)＝exp(β′₁X₁+β′₂X₂+β′₃X₃) (6)

in the formula (6),. beta.'₁-β′₃Alternative path sets C respectively representing users n_nType X of non-motor vehicle lane in each path₁Parking section X in road₂And the number of traffic lights X₃Influence coefficient on riding hazard rate;

obtaining a safety factor S of the user n selection path i after the riding time t by using the formula (7)_niThereby obtaining an alternative path set C_nSafety factor of all paths in (1):

step 5.3, calculating the riding continuity Con:

obtaining the optional path set C by using the formula (8)_nThe riding continuity Con of each path:

logCon＝β″₀+β″₃X₃+β″₅X₅+β″₆X₆+ε (8)

in formula (8), β ″)₀Is a constant term；β″₃、β″₅And beta ″)₆Alternative path sets C respectively representing users n_nNumber of traffic lights in each path X₃Road grade X₅And road surface gradient X₆Influence coefficient on the user riding continuity Con; epsilon is a random utility error of the riding continuity Con;

step 5.4, constructing a correction matrix:

the average evaluation score, the safety factor and the riding continuity degree of each path and the probability of each path being selected form a correction matrix

r_1kRepresenting the probability of the k-th path being selected; r is_2kA user rating score representing a kth path; r is_3kRepresenting the safety factor of the k path; r is_4kRepresenting the riding continuity degree of the k-th path;

step 5.5, calculating the final probability of each path being selected;

determining alternate path set C using equation (9)_nSelecting the final probability of each path, and recommending the probability maximum value as the optimal path to the user;

[p₁ p₂ … p_k]＝[ω₁ ω₂ ω₃ ω₄]·R (9)

in the formula (9), p_kA final probability of being selected for the kth path; omega₁The probability of each path being selected determined before correction is weighted; omega₂Evaluating the occupied weight for the user; omega₃The safety coefficient is the weight; omega₄Weight of riding continuity, and omega₁+ω₂+ω₃+ω₄＝1。

Images