Select Language



Twitter和Tom's Hardware社交媒体中的嗡嗡声预测数据集

Twitter和Tom's Hardware社交媒体中的嗡嗡声预测数据集

328 浏览
0 喜欢
1 次下载
0 条讨论
Computer Regression

This dataset contains two different social networks: Twitter, a micro-blogging platform with exponential growthand extre......

数据结构 ? 33.3M

    Data Structure ?

    * 以上分析是由系统提取分析形成的结果,具体实际数据为准。

    This dataset contains two different social networks: Twitter, a micro-blogging platform with exponential growthand extremely fast dynamics, and Tom’s Hardware, a worldwide forum network focusing on new technology with more conservative dynamics but distinctive features. Tom’s Hardware (TH) and Twitter (TW) have very distinctive properties:

      • Both German and French contributions are found on (TH), while English contribution are also available in the sample of (TH) we have collected

      • In (TW) there is no direct audience estimator, and we use the nad feature as the target feature, while it is given in (TH) by the number of displays, that measures the number of times a content is displayed to visitors

      • (TW) shows a higher reactivity of exchanges than (TH) 80% of re-tweets take place in the day following the initial tweet, whereas in th the replies to a thread are mostly produced during the subsequent week;

    • The (TW) community is broader than the (TH) one, with more than 500 million visitors per month against nearly 41 millions in (TH)

    In this study, we focus on 6671 topics, such as: over-cloaking; grafikkarten; disque dur; android; etc. related to the technology domain. Here is a summary per language.


    This dataset is published using the UCI guidelines. Hence examples are stored using a standard comma separated value (CSV) format. You will find an example per line, and one feature per column. Each dataset is provided with an additional instructions file, as suggested by the UCI.


    In the classification task you will be provided with time-windows showing an upward trend. The objective of this task is to determine whether or not these time-windows are followed by buzz events. In this task:

      • Each example matches an upward window. Such an example is a multivariate time-series ranging from t to t+β.

      • The labeling (ie. buzz; non-buzz) of an example, as well as the upward detection, are performed considering an univariate time-series. This time series (Y, the target feature, presented bellow) is meant to reflect the popularity of a topic.

      • There is two ways to label examples: Absolute labeling and Relative labeling. The first one labels as buzz any example which is “sufficiently” popular between t+β+1 to t+β+δ while the second one is based on the increment of popularity level before and after β

    • For both of these labeling methods, the threshold value σ varies in order to qualify buzz of distinct magnitude. Concretely σ = 500 implies that an example is labeled as a buzz if:

      • (Relative labeling) the difference between (a) the Y’s mean value between t+β+1 and t+β+δ and (b) the Y’s mean value between between t to t+β is greater than 500

      • (Absolute labeling) the sum of Y’s values ranging from t+β+1 to t+β+δ is greater than 500


    As in the classification task you will be provided only with upward-windows. The value to be predicted will be the value of the time-series used to determine the popularity of a topic (Y, as presented before)




    • 分享你的想法


    所需积分:10 去赚积分?
    • 328浏览
    • 1下载
    • 0点赞
    • 收藏
    • 分享