Newsletter No. 416

4 No. 416, 19.4.2013 今 年1月，全球最大的網絡公司谷歌發出預警，流感 即將在美國爆發，全國瀰漫一片緊張氣氛。谷歌的 「流感趨勢」在2008年推出，基本原理是分析累積的搜索 數據來預測流感潮的起落。據「流感趨勢」的網頁所述， 每星期有數以百萬計的人在網絡世界裏搜尋有關健康的 資訊，搜索主題亦隨季節而異，例如夏天特別多人搜索日 曬炙傷的資訊。研究人員發現，搜索流感主題的次數與實 際有流感症狀的人數有密切的關係，利用這些數據的分析 便可預測流感在某地區的爆發情況。「流感趨勢」當然未 能取代現行的疾病監測系統，但推出以來表現理想，甚至 比美國疾病控制與預防中心更快預測到流感的爆發。 計算機科學與工程學系的 金國慶 教授解釋，谷歌的「流感 趨勢」是社會計算的一個應用實例。社會計算簡單來說， 是利用機器學習、數據挖掘和網路智能的技術來分析網 誌、臉書、YouTube等社會媒體所產生的大量數據，從中了 解人們的社會關係。「這對近年愈來愈重要的『大數據』 意義非比尋常，」金教授說。「過去的十年八年間，湧現了 很多新的社會現象，產生了數量驚人的數據，社會計算就 是如何去分析和理解這些數據，現時已成為計算機科學的 一個重要範疇。或者可以說，社會計算是社會行為和計算 系統的交匯。」 「大數據」現時是所有公私企業和機構都不容忽視的課 題。金教授在中大研究和教授的其中一科就是社會計算， 他指出，我們在互聯網上和使用流動通訊器材的所有活動 都成為數據，企業和機構掌握了這些數據，可從中梳理出 顧客或服務對象的喜好和特徵，從而設計更符合市場的產 品或服務。所以企業和機構毫不怠慢，大家的態度都是先 把數據儲存起來再說，如何分析和利用是下一步的事。 有幾個修讀金教授的社會計算課的內地生利用所學，完成 了一項有趣的研究作業。他們分析了南京計程車的全球定 位系統數據，按市內不同地區不同時段的交通情況，找出 最快捷的行車路線，還計算到司機的收入情況，平均是多 少，最高和最低是多少，一目了然。 金教授去年才回到中大。之前的兩年，他於加州大學柏克 萊分校擔任客座教授，講授兩門與社會計算及大數據相 關的課程。同時，他在三藩市的美國電話及電報公司做研 究，從人們的打電話和通訊模式中，推斷一個人的身分， 然後概括出這個人的行為和習慣。這類研究不但有巨大的 商業潛力，在了解人的行為上亦有很大的啟發。金教授亦 不諱言，類似的研究全球不少政府都在進 行，致力應付恐怖活動的國家尤其不敢掉以 輕心。 金教授指出，計算機科學裏的數據分析絕非新鮮事物，但 隨着互聯網和流動通訊技術的發達，數據分析的發展一日 千里，社會計算的興起可以說是一個高峰。他說：「現今的 數碼社會是一個開放的群體。參與其中便要有心理預備， 一舉一動都會留下蹤跡供人追尋，所以適當的安全措施一 定不可少。不過最重要的還是要自己小心言行。保障私隱 畢竟要由自己做起。」 I n January, the world’s largest search engine Google warned of a flu outbreak in the US, causing a ripple of anxiety across the country. Google Flu Trends, a project launched in 2008, analyses cumulative search figures to make forecasts about the ebb and flow of flu trends. According to the Google Flu Trends website, millions the world over search health-related information on the Internet every week, with search topics changing according to the seasons. For instance, in summer, there are more people looking for content on sunburn. Researchers discover that the number of times flu topics are searched corresponds to the number of people exhibiting flu symptoms. Analysing these data allows them to predict a flu outbreak in a certain region. This doesn’t mean that Google Flu Trends can replace a disease monitoring system of course. Nonetheless it has performed laudably since its launch, even beating the Centers for Disease Control and Prevention (CDC) in the US in predicting a flu outbreak. Prof. Irwin King of the Department of Computer Science and Engineering explained that Google Flu Trends is an example of how social computing can be applied to real life. Simply put, social computing is the use of machine learning, data mining and web intelligence to analyse the massive volumes of data generated by blogs, Facebook, YouTube and other social media, in order to gain insight into social relationships. ‘This has unusual implications for “big data” which is of increasing importance,’ Professor King pointed out. ‘The last eight or 10 years have seen the blossoming of many new social phenomena, which has given rise to an incredible amount of data. Social computing which analyses and tries to understand such data has become an important area in computer science. You can say that social computing is where social behaviour meets computing systems.’ ‘Big data’ is a topic public and private corporations and organizations cannot afford to ignore. One of the subjects Professor King studies and teaches at CUHK is social computing. He remarked that all activities we conduct on the Internet via computers or mobile communication equipment become data. When corporations get hold of such data, they can identify trends in the preferences and characteristics of their service targets. This helps them to design products and services that better fit market needs. Not surprisingly, organizations have been quick to get their hands on such data with the attitude often of ‘store first, decide later’. Several mainland students who took Professor King’s social computing course have completed a fascinating project. They analysed GPS data from taxis in Nanjing, and using those on traffic conditions at different time periods in different parts of the city, worked out the most efficient routes. What’s more, they managed to calculate the median, highest and lowest incomes of taxi drivers in the city. Professor King had been a visiting professor at UC Berkeley for two years, returning to CUHK only in 2012. While at Berkeley, he taught two courses related to social computing and big data. He also conducted research at AT&T in San Francisco, making forecasts about a person’s identity and subsequently behaviour and habits based on his or her patterns of telephone usage and communication. This kind of research not only has enormous commercial potential, it is also helpful for understanding human behaviour. Professor King observed that many governments are engaged in similar research, in particular, those grappling with terrorism. He added that data analysis is nothing new in computer science, but with the boom in Internet and mobile telecommunication technologies, the field is developing at an astronomical rate. The rise of social computing can be said to be a peak of such developments. ‘Contemporary digital society is an open group. Participants should be aware that anything they do will leave traces. They must take the necessary safety precautions. More importantly, they should watch what they say or do. Protection of privacy should start with the self.’ Where Social Behaviour Meets Computing Systems 社會行為與  計算系統的交匯