考量社群媒體表情符號與內文之動態再調整投資組合模型

Dynamic Portfolio Model with Considering Sentiments: Analyzing Emoji and Text on Social Media

吳青浩(Ching-Hao Wu)

指導教授 ： 余菁蓉

國立暨南國際大學/管理學院/資訊管理學系/碩士(2018年)

摘要

研究指出社群媒體上的使用者投資理財意見與情緒會直接影響股市走勢，但是至目前為止，並沒有人詳細探討分享文章含表情符號對股市的漲跌是否有影響，也少有研究將情緒分析結合傳統投資組合並進行再調整。因此本研究使用條件風險值模型及Omega 模型，並考量交易成本與放空來做為測試的基礎，衍伸過去固定時間再調整投資組合，透過情緒詞典 Stock Market Lexicon(SML)與表情符號情緒分數(Emoji Sentiment Ranking)來分析使用者對股市的看法，作為動態再調整投資組合啟動的依據。實驗資料為2017年7月5日至2018年3月10日，共249天S&P500 成分股的日資料，動態再調整次數為178次，並且使用CVaR模型與Omega模型，實驗分為五個部分：動態再調整考量與不考量表情符號之績效比較、負向情緒放空規則不同的績效比較、固定天數再調整績效比較、買進與放空權重下限比較與依據情緒分數調整權重下限，最後有交易成本與損失值的探討。本實驗自動化計算社群媒體意見並運用於動態再調整機制，並且我們發現動態再調整的投資組合模型因情緒語意分析不同而給予不同權重時，能夠使投資組合績效提升。

關鍵字

社群媒體 ； 表情符號 ； 結構化資料 ； 非結構化資料 ； 動態配置再調整 ； 交易成本

並列摘要

Nowadays, it’s a very popular issue that user opinions on social media may directly affect the stock market assets. But so far, few studies have discussed whether essay with emoji has an influence on the stock assets. Besides, few studies use the sentiment analysis to solve the bottleneck of the convention portfolio. Therefore, in our study, we not only used linear models (CVaR and Omega) and considered transaction costs and short selling, but also utilize Stock Market Lexicon (SML) and emoji emotional scores (Sentiment of Emoji) to analysis user's perception of the stock assets. Our experimental data is from July 5, 2017, to March 10, 2018, with a total of 249 days of S&P500 constituent stocks and the number of rebalancing is 178. Out experiment can be divided into five parts: Performance comparison between dynamic rebalancing and without emoji, a performance comparison of negative emotion venting rules, fixed-day rebalancing performance comparison, comparison of the lower limit of buy-and-sell weights, lower weight limit based on sentiment scores, and discussion of transaction costs and loss values. In our research, we apply the value of social media in our dynamical rebalance mechanism. Furthermore, by adjusting the weight of sentiment analysis, we can improve the performance of our mechanism.

並列關鍵字

Social media ； emoji ； structured data ； unstructured data ； dynamic configuration re-adjustment ； transaction costs

參考文獻

Artzner, P., Delbaen, F., Eber, J. M., and Heath, D. (1999). Coherent measures of risk. Mathematical Finance, 9, 203-228.

Baccianella, S., Esuli, A., and Sebastiani, F. (2010). SentiWordNet 3.0: An enhanced lexical resource for sentiment analysis and opinion mining. In Proceedings of the Seventh Conference on International Language Resources and Evaluation LREC10 (pp. 2200–2204).

Bardong, F., Bartram, S. M., Yadav, P. K. (2008). Are short-sellers different? EFA 2008 Athens Meetings paper, Retrieved from http://ssrn.com/abstract=1101786

Beraldi, P., Violi, A., and De Simone, F. (2011). A decision support system for strategic asset allocation. Decision Support Systems, 51, 549-561.

Cogent Research, 2008, “Social Media’s Impact on Personal Finance & Investing.” Available from: http://www.cogentresearch.com.