CN107122481A - News temperature real-time online Forecasting Methodology - Google Patents

Abstract

The invention discloses a real-time online prediction method for news popularity, which includes two parts, hot event analysis and modeling and latest news popularity prediction, and hot words and hot word pairs in all events are combined to form hot words and hot word pairs The calorific value table of hot words in each event is added to obtain the current calorific value of hot words and hot word pairs; the above hot words and hot word pairs are constantly updated; The calorific value table of the hot words and hot word pairs obtained in the modeling step is used to score the vocabulary and vocabulary combinations obtained in the calorific value table, that is, the calorific value of the vocabulary and the calorific value of the vocabulary combination are queried in the calorific value table, and the same vocabulary and vocabulary The heat value of the combination is accumulated to obtain the heat value of each word and word combination in the current news; the heat value of all words and word combinations in the news is added to obtain the news heat, which is the predicted news heat. The invention can comprehensively analyze hot topics and update hot news in time.

Description

Real-time Online Prediction Method of News Popularity

technical field

The invention relates to the field of news information, in particular to a method for real-time online prediction of news popularity.

Background technique

With the rapid development of Internet technology, Internet public opinion is increasingly affecting the stable development of society. Monitoring Internet public opinion is an important link for the government to maintain social stability. As one of the links in public opinion monitoring, the prediction of hot news is particularly critical. With its unique dissemination characteristics and real-time interaction characteristics, Weibo has changed the way of dissemination of traditional news information. In particular, the combination of Weibo and mobile terminals enables Weibo information to be forwarded or commented on more quickly, and a large number of user comments and exchanged information on the Weibo platform can be quickly gathered into opinions, thereby forming a certain trend of public opinion. Weibo's natural openness, real-time, interactive, massive and easy-to-check nature constitute the basis for hot news forecasting. Judging the popularity of news by comprehensively analyzing the topic volume of news on the Weibo platform.

Traditional public opinion hot topics are judged only by data such as number of clicks, forwarding numbers, and comments, but this hot topic prediction technology cannot fully analyze the characteristics of hot topics, and the extraction of hot news is not timely enough.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the prior art and provide a real-time online prediction method for news popularity, which can comprehensively analyze hot topics and update hot news in time.

The purpose of the present invention is achieved through the following technical solutions:

A real-time online prediction method for news popularity, including the following two parts:

Analysis and modeling of hotspot events, including the following steps:

S01: For the hot events that have occurred, manually determine the keywords, and based on the manually determined keywords, crawl various information related to the hot events from the Internet;

S02: Evaluation of event heat value, using the total amount of information crawled from the web to score the heat of the event. The larger the total amount of information, the higher the score, and there is no upper limit;

S03: Analyze hot words on the crawled information, and find out 20% of the most popular words;

S04: Perform hot event modeling on known hot events, analyze the contribution rate of various entries to the popularity, and the joint contribution rate of the combination of entries to the event popularity;

S05: Use the popularity value of the event, the contribution rate of the hot word to the event, and the contribution rate of the hot word combination to the event to calculate the heat value of the hot word. The calculation formula is: hot word heat value = event heat value × frequency of hot words / all The sum of the frequency of the hot words; the heat value of the hot word pair = the heat value of the event × the frequency of the hot word pair / the sum of the frequencies of all the hot word pairs;

S06: Combine the hot words and hot word pairs in all events to form a hot word and hot word pair heat value table, add the hot words in each event to get the current hot word and hot word pair calorific value;

S07: constantly updating the calorific value table of the above-mentioned hot words and hot word pairs;

The hottest prediction of the latest news includes the following steps:

S11: Collect information from various sources in real time, including but not limited to news, Weibo, forums, and post bars;

S12: Segment the information collected above, remove stop words, and obtain news-related vocabulary;

S13: Use the heat value table of hot words and hot word pairs obtained in the hot event analysis and modeling step to score the words and word combinations obtained in the heat value table, that is, query the heat value of words and the number of word combinations in the heat value table Calorific value, adding up the calorific value of the same vocabulary and vocabulary combination to obtain the calorific value of each vocabulary and vocabulary combination in the current news;

S14: Add the popularity of all words and word combinations in the news to obtain the news popularity, which is the predicted news popularity.

Further, the network in the step S01 includes articles, microblogs, and WeChat content containing the keyword in various channels such as major news websites, Weibo, WeChat, forums, post bars, and government websites.

Further, the formula for calculating the heat value of an event in step S02 is Hotvalue=sum[count×k], where count represents (the total number of public opinions), and k is a weight whose value ranges from 1 to 100.

Further, the hot word analysis in the step S03 includes the following steps. First, stop words need to be removed, and then the vocabulary is scored by frequency of occurrence. Word frequency refers to the number of times the vocabulary appears in all content, and the popularity can be found according to the score. Top 20% of vocabulary.

The beneficial effects of the present invention are: the present invention analyzes the current hot events, scores and sorts out the hot events to form a hot event table, and then scores the collected hot words according to the hot event table, and through a series of calculations Get the corresponding heat value, so as to realize the comprehensive real-time analysis of current hot events and timely update news hotspots.

detailed description

A real-time online prediction method for news popularity, including the following two parts:

Analysis and modeling of hotspot events, including the following steps:

S01: For the hot events that have happened, manually determine the keywords, and based on the manually determined keywords, crawl various information related to the hot events from the Internet; including major news websites, Weibo, WeChat, forums, post bars, government Different channels such as the website contain articles, Weibo, WeChat, etc. that contain the keyword.

S02: Evaluation of event heat value, using the total amount of information crawled from the network to score the heat of the event. The larger the total amount of information, the higher the score, and there is no upper limit; different information sources can manually set the weight, The more important the source of information, the higher the weight of the impact on the score, which can be continuously adjusted according to the business scenario. For example, for news events, it can be set higher if it comes from a website with high credibility, such as People's Daily Online, Xinhuanet, etc. The weight of , for entertainment events, comes from the Weibo information of celebrities and big Vs, and can be given a higher score.

S03: Perform hot word analysis on the crawled information to find out 20% of the most popular words; hot word analysis, first need to remove stop words, and then use frequency of occurrence to score words, word frequency refers to the word appears in all content The number of times, according to the score to find the most popular 20% of the vocabulary, for example, in a certain sports news, "football" appears in the article 10 times, then the word frequency value of the word "football" is 10, all under this topic The word frequency value of "football" in the news is added, and it ranks in the top 20% of the word frequency of all words, then "football" is one of the hot words under the topic.

S04: Perform hot event modeling on known hot events, analyze the contribution rate of various entries to the popularity, and the joint contribution rate of the combination of entries to the event popularity;

S05: Use the popularity value of the event, the contribution rate of the hot word to the event, and the contribution rate of the hot word combination to the event to calculate the heat value of the hot word. The calculation formula is: hot word heat value = event heat value × frequency of hot words / all The sum of the frequency of the hot words; the heat value of the hot word pair = the heat value of the event × the frequency of the hot word pair / the sum of the frequencies of all the hot word pairs;

S06: Combine the hot words and hot word pairs in all events to form a hot word and hot word pair heat value table, add the hot words in each event to get the current hot word and hot word pair calorific value;

S07: constantly updating the calorific value table of the above-mentioned hot words and hot word pairs;

Further, the formula for calculating the heat value of an event in step S02 is Hotvalue=sum[count×k], where count represents (the total number of public opinions), and k is a weight whose value ranges from 1 to 100, such as People's Network, Xinhuanet The k-value of news from websites with equal weight is 100, and the k-value of ordinary user news from we-media channels such as Weibo is 1.

The hottest prediction of the latest news includes the following steps:

S11: Collect information from various sources in real time, including but not limited to news, Weibo, forums, and post bars;

S12: Segment the information collected above, remove stop words, and obtain news-related vocabulary;

S13: Use the heat value table of hot words and hot word pairs obtained in the hot event analysis and modeling step to score the words and word combinations obtained in the heat value. The calculation formula is Hotvalue=sum[count×k], where count means (the total number of public opinions), k is the weight and its value is 1~100. That is, query the calorific value of the vocabulary and the calorific value of the vocabulary combination in the calorific value table, accumulate the calorific value of the same vocabulary and vocabulary combination, and obtain the calorific value of each vocabulary and vocabulary combination in the current news;

S14: Add the popularity of all words and word combinations in the news to obtain the news popularity, which is the predicted news popularity.

The above descriptions are only preferred embodiments of the present invention, and it should be understood that the present invention is not limited to the forms disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and Modifications can be made within the scope of the ideas described herein, by virtue of the above teachings or skill or knowledge in the relevant art. However, changes and changes made by those skilled in the art do not depart from the spirit and scope of the present invention, and should all be within the protection scope of the appended claims of the present invention.

Claims (4)

1. a kind of news temperature real-time online Forecasting Methodology, it is characterised in that including following two large divisions：

Focus incident is analyzed with modeling, and is comprised the following steps：

S01：To the focus incident occurred, keyword is manually determined, based on the keyword manually determined, is crawled from network various The related information of the focus incident；

S02：Event hot value is assessed, the informational capacity crawled using network, and the temperature to event is given a mark, and informational capacity is bigger , score value is higher, no ceiling；

S03：Hot word analysis is carried out to the information crawled, 20% temperature highest vocabulary is found out；

S04：Focus incident modeling is carried out to known focus incident, contribution rate of the various entries to temperature, and entry is analyzed Combine the joint contribution rate to event temperature；

S05：Heat is calculated to the contribution rate of event using the contribution rate of event hot value and hot word to event, and hot word combination The calorific value of word, calculation formula is：The frequency sum of hot word calorific value=event calorific value × hot word frequency/all hot words；Hot word is to calorific value The frequency sum of=event calorific value × hot word to frequency/all hot words pair；

S06：The calorific value table that hot word and hot word in all events to be formed to hot word and hot word pair to being merged together, hot word is existed Calorific value in each event is added, and obtains the current calorific value of hot word and hot word pair；

S07：Constantly update the calorific value table of above-mentioned hot word and hot word pair；

Latest news temperature is predicted, is comprised the following steps：

S11：The information in various sources, including but not limited to news, microblogging, forum, the content of mhkc are gathered in real time；

S12：Participle is carried out to the above-mentioned information collected, removes stop words, the related vocabulary of news is obtained；

S13：The vocabulary that the calorific value table centering for analyzing hot word and hot word pair with being obtained in modeling procedure using focus incident is obtained Temperature marking is carried out with word combination, i.e., the calorific value of vocabulary and the calorific value of word combination are inquired about in calorific value table, by identical vocabulary Added up with the calorific value of word combination, obtain the calorific value of each vocabulary and word combination in Present News；

S14：All vocabulary in news are added with the temperature of word combination, news temperature is obtained, this temperature is the new of prediction Hear temperature.

2. a kind of news temperature real-time online Forecasting Methodology according to claim 1, it is characterised in that：Described step Network in S01 includes major news websites, microblogging, wechat, forum, mhkc, and the different channels of government website include the pass The article of keyword, microblogging, the content of wechat.

3. a kind of news temperature real-time online Forecasting Methodology according to claim 1, it is characterised in that：Described step The calculation formula of event hot value is Hotvalue=sum [count × k] in S02, and wherein count is represented（Public sentiment sum）, k is Its value of weight is 1 ~ 100.

4. a kind of news temperature real-time online Forecasting Methodology according to claim 1, it is characterised in that：Described step Hot word analysis comprises the following steps in S03, it is necessary first to remove stop words, and then vocabulary is given a mark using the frequency of occurrences, Word frequency refers to the number of times that the vocabulary occurs in all the elements, and the vocabulary for finding out temperature highest 20% is settled by score.