2.1 Background on the Orienteering Problem

In the case of orienteering problems, the various control spots with the associated scores are placed at many places [²³]. Competition members strive to maximize their overall performance in the shortest time achievable by reaching as many controls as possible. The most important thing is to reach the highest rating despite the short lifespan [⁸, ²⁸].

2.2 Tour Recommendation based on Orienteering Problem and its Variants

Lim et al. [¹¹] changed the tourism orienteering problems according to the importance of the POI tour guidance model. Vansteenwegen et al. [²⁷] proposed an approach for adapting the tour schedule so that it would improve the overall balance between the defined degree of involvement from the starting and end, such as expenditure and all POIs. Lim and al. [¹²] have identified places of importance and reputation in the form of the minimum queuing time.

Tours that satisfy the different levels of tourist interest within the group have been developed in such a manner as the concept of orienteering problem [², ¹³].

2.3 Different Tourism Related Work

The time concerned measurement technique that considers tourist visitation at an attraction was proposed by Ying et al. [³⁰]. Furthermore, over a while, the result was shown to make this technique complicated. Aliannejadi et al. [¹] developed a possible model to assess the connection between the tourist comments label and a similar attraction. The findings were measuring in combination with studying to rank techniques from different LBSN tools. Given the considerable time, Zhao et al. [³¹] proposed a latent spatial time model to propose optimal subsequent destinations.

Li et al. [¹⁰] recommended that all information from spatial and temporal inspections be stored using the Time-aware Factorizing Personalized Markov Chain (TA-FPMC). The study analyzed the time-decay factor by comparing the gap across two concurrent experiments. Both calculations were built on both check-in experiences and the design was highly complex. The authors also consider the customer as an extra layer that is not necessary for this work. Consequently, if included in the recommended framework, the productivity of a future customer will decrease.

Our suggested model differs significantly from the current POI and tour recommendation scheme: Our algorithms categorize the interests of visitors dynamically depending on time and popularity with geo-tagged images. The POI tourist costs and local distance between the past POI of the initial path and the initial POI of the subsequent path are also reduced in consideration.

If a tourist goes to a new geographical area without a history of his experience, the suggested methodology can offer a suggestion. While some of the solutions suggested have a similar viewpoint, those only have one route in unexplored locations, depending on the kind of POI he/she has experienced The existing approach discovers the connection between a familiar and an unfamiliar location.

The proposed approach recommends several POIs and the associated POIs are also coordinated.

3.1 Average POI Visit Duration for Local and Global Users

Every tourists U are aware of the tour’s past information. In a specific POI, Eqn. 1 can be used to measure the duration:

where J={1,2,…,m˙}, u={1,2,…,k′} and ∇ and denotes the number of travels to a given POI by δ(Pj=P)={​0, otherwise1 if (Pj=P). In the case of all tourists, Ξ(P) is often employed [⁴, ⁷].

3.2 Time-based user interest for local and global users

ℂP represents POI group P, as shown in the prior section. Eqn. 2 helps to register the interest of a particular tourist u˙ in POI group c˙:

where δ(CPJ)={​0, otherwise1 if CPJ=c˙.

The tourist interest for the POI group c˙ is measured by Eqn. 2 as per the spending time on the POI group c˙ based on the total spending time by all the tourists. It seems obvious that the tourist will devote most of his or her time at that POI.

3.3 Similarity of Local and Global Users

The identity of local and global tourists are computed based on the Cosine Similarity measure based on the interest of a given destination for both local and global tourists and which is computed utilizing Eqn. 3:

where u˙x˙ and u˙y˙ are the 2 distinct tourists.

3.4 Itinerary from Travel History

The traveling reports have been defined for a particular tourist u˙∈U , based on the sequence n˙ travel POIs Su˙=((P1,tP1a,tP1d),…,(Pn˙,tPn˙a,tPn˙d)), wherein a triplet (Py˙,tPy˙a,tPy˙d), where Py˙ is the tourist’s traveled POI, tPy˙a and tPy˙d are the time of entry and exit.

The difference between the time of entry and exit indicates the duration of the POI Py˙. Here, Su=((P1,tP1a,tP1d),…,(Pn˙,tPn˙a,tPn˙d)) could be re-written as Su˙=(P1,…,Pn˙).

3.5 Time-based user Interest of a POI

The POI interest Py˙ is a part of Su˙ and could be calculated with the help of Eqn. 4:

3.6 Popularity of a POI Category

The POI popularity is measured by using Eqn. 5 as per the overall number of tourists visiting on the POI with respect to to the number of tourists visits to every POI:

where the ℂ(pop¯) represents category ℂ of popularity and φ(P) represents the number of instances of all tourists visiting a specific POI.

3.7 Traveling Cost

The cost of travel is determined using the actual path the traveler travels through. While a few previous studies have taken into account the full length of the trip, distances are a relevant consideration for the tourist recommendation that the costs trigger when a tourist selects a long-distance pattern to visit different POIs.

We will minimize traveling time by utilizing quicker modes of transportation. If the gap between the two POIs increases and in that situation traveling costs will therefore arise, a quicker form of travel is necessary. We however intended to minimize the length of the tour. The transportation costs are calculated by Eqn.6:

The first portion of Eqn. 6 consists of the inner length of all POIs in the pac plan. The inner length of the POI Py˙ is calculated by the total length of all POIs. The second portion of Eqn. 6 reflects the consecutive real length among Py˙ and Py˙+1 POIs, which could be calculated by taking account of the length between the two consecutive POIs.

3.8 Problem Definition

This portion deals with the problem of different POIs for one person. Our main objective is to maximize the interest and popularity of visitors and to reduce expenditures. A type of orienteering issue [¹²] could be used to resolve this issue:

Our aim here is to propose an itinerary focused on tourist interest for specific POI , tour popularity and weather interest. The weight parameter can be adjusted as needed. The key goal of this arch is to propose an itinerary to increase the interest, popularity of visitors, weather interest and minimize travel expenses. The user’s interest in various weather conditions, such as winter, summer, and so on, is used to quantify the weather interest, which is denoted by W(inte). For eg, if a visitor prefers to visit a location in the winter, this indicates that the atmosphere is appealing to him.

The aim is to find an itinerary tour plan I=(P1,P2…,Pn˙) that:

Let TP,P′=1, if the traveler has explored the POIs P and P′ sequentially. That means a tourist can travel from P to a P′. TP,P′=0, otherwise [¹¹]. Then Eqn. 8 could be overcome with 19-22 restrictions:

In Eqn. 8, the multi-objective issue is solved by increased visitor interest and popularity, weather interest and a decreased expense. The limitation set out in Eqn. 9 means that the proposed plan should be started from the first POI and the last POI can be completed. The limitation provided in Eqn. 10 shows that there is no POI viewed multiple times.

The restrictions imposed in Eqn. 11 and 12 argue that another response would not include a sub-tour based on the issue of the traveler with the sub-tour eliminating problem 13. Section 15 guarantees that the overall gap for the kit is provided by budget B. Li′ represents the i′-th itinerary herein.

The issue is an NP-hard problem since it relies on the cost function Cost(P1…Pn˙). This is also affected by the multiple POIs selected from a wide variety of variants. To address these problems, we are proposing the wPOIapproach is dependent on Monte-Carlo Tree Search (MCTS) and is described herein.

3.9 Monte Carlo Tree Search Algorithm

The Monte Carlo Tree Search (MCTS) method is used in board games including Othello, Chess, and Go [⁵, ²⁸]. The MCTS algorithms are based on the tree search idea. All the boards in the board graph are referred to as the node and the game’s result score is considered the leaf node. The gameplay can take multiple runs to hit the endpoint based on the MCTS formula.

Any execution starts with a set of randomized nodes and reports the results. The following steps are performed to get the winning score. A set of trials have been conducted in MCTS (e.g. 100 trials) and it is repeated for a specific duration ( e.g. 10 seconds). The below are the basic core tasks of MCTS:

MCTS solution is used for various networking problems, such as the issue of travel salesmen [¹⁸] or car navigation [⁹]. This issue is fixed based on two main factors by the use of the MCTS approach [⁵].

Due to the below reasons – MCTS should not be directly implemented:

The Upper Confidence Bound (UCT) is used to navigate the POI Pn˙, which maximizes Eqn. 14:

The initial UCT (Eqn. 14) is the improvement of our suggested technique where a probabilistic selection is made for the succeeding POI and is described in Eqn. 15.

where VCount denotes is the number of visits of viewed nodes and TReward Reward is the cumulative reward from proposed POIs.

3.10 Simulation and Back-Propagation

Inside the MCTS, the test started at the root node, evaluating the reward as 1 (win) and 0 (loss). For binary 1 and 0 numbers, the reward for some POIs is not fully defined. The reward depending, on various parameters including tourist interest, tour popularity, and travel expenses, and is described as:

For each loop, the reward score is computed by Eqn. 16. If the itinerary is successful, the reward score will be back-propagated to all viewed nodes Furthermore, the number of trips is replicated and then raised by one.

4.1 Dataset

In this analysis, we utilized the data provided in [¹²]. The dataset contains images and videos by Yahoo! Flickr Creative Commons 100M (YFCC100M) [²⁵]. Furthermore, the YFCC100M data set provided in [¹²] was used and geo-tagged images from different areas of the globe have been obtained. The data set comprises the photo’s meta-data. It includes visiting dates and times. The dataset also contains data from the Geo-coordinate to identify the length among POIs. The data sets utilized in this research could be accessed^fn freely.

4.2 Baseline Algorithms

Based on the work [¹²] we have taken into account all the benchmark algorithms beginning at one POI and then choosing the next following POI before the budget is achieved. We utilize the tour series by the tourist to suggest multiple POIs.

4.3 Real-life Evaluation

Only visitors who have completed at least two travel sequences and two groups are assessed. The method is applied to both local and global datasets [²⁴], as well as visitors who are comparable. We compare similar visitors in this study by looking at the top 10 associated visitors from global data sets. To equate different baselines with our method, we chose the preceding formulas. For our experiments, categories of real traveling series are chosen based on the history of associated visitors in a given area.