Teach you how to understand travel big data.

Teach you how to understand travel big data _ data analyst exam

Sometimes, in a word, a picture will contain huge digital business opportunities, but this is a technical job with high precision, and not everyone can understand big data.

Understand the behavior of tourists

Everyone is talking about big data. Ctrip recently invested in Zhonghui Information Technology Co., Ltd. (hereinafter referred to as "Zhonghui"), Ali Travel cooperated with Shiji Information, and Dongcheng Hotel and Home Inns also launched intelligent management.

Every visitor will have his own membership and consumption data records, which are the basic information of big data. However, how to analyze and deal with a bunch of numbers and consumer behavior and draw a conclusion?

"The first thing to know is what big data is. Big data is divided into two categories, namely structured data and unstructured data. The former is a series of numbers you see, and the latter may be a picture, a sentence, etc. Information that is not directly reflected as numbers. Therefore, real big data analysis should not only do direct digital analysis, but also know how to build mathematical models, transform unstructured data into structured data and draw conclusions. These are not simple. " Jiao Yu, general manager of Zhonghui Data Intelligence Division, told the reporter.

Jiao Yu gave the reporter an example. Now many tourists will compare prices and book hotels on OTA (Online Travel Agency), so their search keywords and browsing traces will be reflected in the records of OTA. If a guest browses the page of this hotel but jumps and doesn't place an order, you can analyze the reasons why the guest didn't place an order through this record. When the guest sorts the hotel information by keywords such as price, brand and region, the browsing records he left can show that people are more sensitive to price.

"After research, most people still pay attention to the price factor. Because the price choice is interval, we can use the browsing trace to get a price interval map that is most acceptable to tourists. Only 1 1% people care about the brand, which shows that similar hotels are highly replaceable. If you search by regional keyword, it represents geographic location data. If we can be accurate to the specific location and convey this information to the hotels in this area, it will undoubtedly increase the hotel occupancy rate and adjust the house price appropriately according to consumer behavior. When the supply exceeds the demand, the house price will decrease, on the contrary, the house price will increase. There is also a very interesting study, that is, if there is a competitor hotel of Hotel A in the tourist browsing record, it can be inferred that this guest has a demand for Hotel A, and this guest is a potential guest that Hotel A should pay attention to. " Jiao Yu pointed out that in order to transform the massive browsing records into effective data, we have to rely on mathematical models, which can be divided into convergent and divergent types. Big data usually needs to transform unstructured data into structured data through a convergence model and draw conclusions.

A chain hotel operator told reporters that these big data cover information such as price range and brand acceptable to consumers, which can make strategic adjustments in price, positioning and marketing of hotels, thus increasing occupancy rate and improving overall revenue management of hotels.

Magical language analysis

In addition to price and brand, language is also a kind of unstructured data. Especially now, when guests book hotel tourism products, they will definitely read the comments first, or experience them themselves before leaving comments. There is also a lot of big data knowledge behind these languages.

After many interviews and observations, the reporter learned that many guests will evaluate the hotels they have stayed in. Among these comments, there are often comments on the hotel environment, room facilities, catering and service, such as "the room is clean, but the food delivery service is slow", "the front desk service is poor" and "it feels good to take a shower". This requires professional semantic analysis, which accurately subdivides it into structured data and feeds it back to hotel operators.

In artificial intelligence and computational linguistics, semantic analysis provides a method for knowledge reasoning and language, and it is also the development direction of future search engines. For example, by inputting "Apple" through semantic analysis, we can know that users are looking for mobile phones instead of fruits.

"First of all, we will remove a batch of false comments or comments without substantive content through professional semantic analysis, and leave comments with real substantive content to the hotel, and break sentences and cut each sentence in multiple dimensions. For a simple example, for example,' This hotel is clean, but the food delivery service is slow', after our sentence segmentation and multidimensional cutting analysis, we can know that the cleanliness of the room is good, but there is something wrong with the food delivery, so we will classify the conclusions in detail and feed them back to various departments. The problem here is speed, and sometimes it involves taste or service attitude. Sometimes the analysis of a paragraph is very complicated, including the proportion of error correction. " Fan Hu, senior vice president of Zhonghui Marketing Department, said.

Mr. Li, who has been engaged in the hotel industry for more than 15 years, told reporters that compared with simple "good" or "poor", the conclusions drawn after multi-dimensional semantic analysis can be fed back to all relevant departments of the hotel, and the details can be refined. Good, or what details need to be improved, then the management can clearly know the next work direction when holding a regular meeting, and after improving the service attitude, the speed and even the decoration style, The occupancy rate of the hotel where it is located has increased by 10%, and RevPAR (the average actual operating income generated by each room available for rent) has increased by about 15%.

It is reported that the dimension of semantic analysis of some science and technology information companies has reached 1000.

How to play cross-border and picture information

Sometimes, the analysis of tourism big data will also involve cross-border cooperation.

"Foreign countries are interdisciplinary research, combining many fields, such as geographic information, IT, business schools, sociology and so on. Let me give you an example of tracking tourists. Now we use cross-border multi-directional social media to track the behavior of tourists. Tourists leave many traces on social media, such as flickr, and the pictures on flickr leave geographical coordinates, shooting time, comment information and so on. These are very valuable travel big data. "Cheng Mingming, a scholar who has been studying tourism big data analysis in Australia for a long time, told reporters that tracking the trajectory with geographical coordinates requires the help of experts who understand geography, while experts in business management can analyze where tourists go and when to go. Data with commercial value.

After cross-border analysis and research, the industry can know which scenic spots are popular, which scenic spots are relatively new, and the time for tourists to stay at the scenic spots or even the time for each stay. After mastering the analysis results of these big data information, the relevant tourism industries can effectively divert the flow, and will not cause the carrying capacity of scenic spots to be too saturated. At the same time, by comparing the information of scenic spots with the attributes of tourists, we can know what different needs tourists from different countries have for scenic spots, such as whether Asians prefer cultural attractions, and if so, more cultural attractions should be introduced into local tourism promotion and marketing.

In the interview, the reporter learned that many scenic spots in China are also cooperating with relevant big data analysis companies, hoping to predict the future passenger flow through analysis, especially during the Golden Week in the peak season, which can help scenic spots control the number of tourists and improve safety and service quality.

Interestingly, pictures also belong to big data.

"For example, there are a lot of pictures on some large travel booking websites. For pictures, we need IT technicians to help machine learning to help us identify different pictures. For example, whether the characters or scenery are good or not, then we compare the mathematical model with the pictures promoted by tourism bureaus and travel agencies to find out whether the pictures that tourists are interested in are consistent with those promoted by tourism bureaus and travel agencies. If it is inconsistent, what is the inconsistency and how to improve it. " Cheng Mingming said.

It is reported that there is also a brain wave test method, which can test where people will look at a picture in the first second, that is, the most attractive point, and the degree to which people like or dislike the picture under test. Through these analyses, the operator can decide whether to change the sample map when selling, whether the promotional pictures of restaurants or scenic spots are good or empty, and the appropriate sample map can promote sales.

"Of course, it is not easy to do a good job in tourism big data research. Its mathematical model is more complicated, such as linear regression. In fact, big data research is a continuous data integration and interdisciplinary process, and there are still many business opportunities for big data in the future. " Cheng Mingming said.

The above is what Bian Xiao shared for you about teaching you how to understand tourism big data. For more information, you can pay attention to Global Ivy and share more dry goods.