【摘要】：隨著金融大數(shù)據(jù)技術(shù)的發(fā)展和投資者個性化、多樣化的需求,金融知識服務(wù)面臨著更大的挑戰(zhàn)。投資者如何獲取及時準(zhǔn)確的實(shí)時數(shù)據(jù),如何估計(jì)某一只債券的未來價(jià)格和未來收益,如何在收益相當(dāng)?shù)膫羞x擇風(fēng)險(xiǎn)最低的債券,如何在最短的時間內(nèi)使得收益最大化等等,都是投資者關(guān)心的熱點(diǎn)問題。以上問題亟需異構(gòu)信息處理技術(shù)、數(shù)據(jù)挖掘方法的技術(shù)支持,因此,本文主要研究從海量金融信息中發(fā)現(xiàn)、挖掘出更有價(jià)值信息的方法和策略,應(yīng)用于企業(yè)債券知識服務(wù)。債券信息的獲取是知識服務(wù)的基礎(chǔ)。為了保證知識服務(wù)的準(zhǔn)確性和高效性,首先需要獲取全面而準(zhǔn)確的數(shù)據(jù),其次,通過進(jìn)一步去噪、優(yōu)化等處理,將金融數(shù)據(jù)處理為結(jié)構(gòu)化數(shù)據(jù),為整個服務(wù)過程中的推薦策略和趨勢預(yù)測提供準(zhǔn)確的數(shù)據(jù)保障。為了使投資者投入更少的精力而獲得相對較高的收益,提出具有針對性的同類益高債券推薦策略。從債券投資者的角度出發(fā),深入分析、研究了投資過程中影響債券收益率的關(guān)鍵特征組合,從而為用戶提供更高效的、個性化的投資策略。債券趨勢預(yù)測為投資者提供了債券價(jià)格和收益變化趨勢的參考。利用機(jī)器學(xué)習(xí)的方法基于債券價(jià)格時間序列、行業(yè)、公司新聞等信息對債券未來趨勢進(jìn)行預(yù)測,綜合多個影響債券價(jià)格走勢的特征因素及多種特征形式,提高了預(yù)測的準(zhǔn)確性。綜上所述,本文利用自然語言處理技術(shù)、數(shù)據(jù)挖掘方法從海量金融數(shù)據(jù)中獲取數(shù)據(jù)信息,并處理為結(jié)構(gòu)化數(shù)據(jù),進(jìn)一步發(fā)現(xiàn)、挖掘更有價(jià)值的信息,利用債券推薦策略和趨勢預(yù)測方法為用戶提供個性化、多樣化且高效的金融知識服務(wù)。
[Abstract]:With the development of financial big data technology and the individualized and diversified demand of investors, financial knowledge service is facing more challenges. How to obtain timely and accurate real-time data, how to estimate the future price and future income of a certain bond, how to choose the lowest risk bond in the equivalent bond, how to maximize the return in the shortest time, etc. Investors are concerned about hot issues. These problems need the technical support of heterogeneous information processing technology and data mining method. Therefore, this paper mainly studies the methods and strategies of mining more valuable information from the massive financial information, and applies them to corporate bond knowledge services. The acquisition of bond information is the basis of knowledge service. In order to ensure the accuracy and efficiency of knowledge service, first of all, we need to obtain comprehensive and accurate data. Secondly, through further de-noising, optimization and other processing, the financial data is processed into structured data. Provides the accurate data guarantee for the recommendation strategy and the trend forecast in the whole service process. In order to make investors invest less energy and obtain relatively high returns, the paper puts forward the recommendation strategy of the same kind of higher interest bond. From the point of view of bond investors, this paper analyzes the key characteristics of bond yield in the process of investment, and provides users with more efficient and personalized investment strategies. Bond trend forecast provides investors with reference to bond price and yield trends. The method of machine learning is used to predict the future trend of bond based on the time series of bond price, industry, company news and so on. The accuracy of prediction is improved by synthesizing many characteristic factors and various characteristic forms that affect the trend of bond price. To sum up, this paper uses natural language processing technology, data mining method from massive financial data to obtain data information, and processing as structured data, and further discover, mining more valuable information, Using bond recommendation strategies and trend forecasting methods to provide users with personalized, diversified and efficient financial knowledge services.
【學(xué)位授予單位】：哈爾濱工業(yè)大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：F830.91;TP311.13

【參考文獻(xiàn)】

相關(guān)期刊論文 前1條

1 陳椺;王雷;蔣子云;;基于K-prototypes的混合屬性數(shù)據(jù)聚類算法[J];計(jì)算機(jī)應(yīng)用;2010年08期

，

本文編號：2245371