众筹项目推荐:面向隐式反馈的深度学习协同过滤

目前,在推荐系统研究中,用户的隐式反馈,以及极度稀疏的数据,已成为影响协同过滤推荐效果的主要问题.针对这一现象,本文提出了深度学习协同过滤算法,先利用卷积神经网络,对用户-项目矩阵的隐层特征进行学习,再结合协同过滤,对用户-项目的交互信息进行建模,并将两种特征融合预测推荐列表.以众筹平台的数据为实验对象,比较模型中各参数对推荐效果的影响,并设计与基线方法的对比实验.实验结果表明:均匀采集负反馈,并在一定卷积层数的网络中,数据稀疏度越高,效果越好;对比基线方法,本文提出的算法在公开数据集(Yahoo!Movie)上取得了最好的推荐结果.本文提出的算法有助于提高众筹平台的融资成功率,同时也丰富了推荐系统的研究体系.     

众筹融资平台

随着互联网金融的热潮兴起,众筹平台也开始成为普通互联网用户关注的热点。事实上,众筹平台要解决的最重要的问题在于信任,只要投融资双方足够信任,那么连淘宝也可以成为众筹平台。而随着众筹平台上出现的～些问题,绝大部分众筹平台都表示将不对平台上出现的项目真实性负责。     

投资者视角下的奖励型众筹问题研究

随着众筹行业的迅猛发展,众筹项目数量迅速增长,使得投资者在项目选择上花费了大量的时间精力.本文旨在帮助投资者以最少时间成本选择优质的众筹项目.在假设众筹项目优质程度与融资完成比有正相关关系的前提下,本文基于京东众筹数据,利用CART回归树算法进行决策树建模,模型R²达到0.746.研究结果表明,投资者应重点关注目标金额,关注人数,项目进展和话题这四个指标.本文研究结果仅适用于奖励型众筹,对于其他类型众筹应当重新选择自变量进行模型建立,但决策树模型仍然可以适用.     

众筹游戏进入冰河期 募资项目大减

众筹,全称大众筹资或群众筹资,是近年来兴起的一种电子游戏开发新模式,游戏开发者通过在网络平台发布自己的游戏项目向大众募集资金,而玩家通过给该项目投资的方式获得将来的游戏测试资格、游戏正式版以及其他一些特权。然而通过全球著名众筹平台Kickstarter的最新一份报告显示,电子游戏的众筹正进入冰河期。     

国内众筹网站举步维艰

国内的众筹网站不是仅仅从美国复制一个模式就能成功的。如何能走出一条适合中国国情的众筹之路,言之尚早。众筹,顾名思义,就是向大众筹资。它利用互联网的传播特性,向广大网友募集资金。相比于传统的融资方式,众筹门槛低,精髓在于小额和大量,不再以是否拥有商业价值作为唯一的评判标准。可以为新型创业公司融资提供了一种新的思维——不是非"大投行、大PE、大VC"     

被误解的众筹

众筹模式从2011年在国内出现开始,一直默默无闻。近日,随着互联网金融大热,被互联网金融从业者们极度推崇的众筹模式逐渐进入公众视线。一些从业者们向记者勾勒出了他们心目中的众筹一众人筹资、众人分红的类P2p模式。从字面来理解,众筹即众人筹资,从展现形式上也与筹资密不可分,在国内现有的法律政策背景下,人们总会有意无意地把众筹与非法集资联系在一起。事实上,有关非法集资的司法解释更多的是针对线下的经济活动制定,线上集资行为到目前为止仍然是监管的空白。     

融合用户兴趣表征与注意力机制的推荐算法

为利用用户行为挖掘用户的兴趣,提出一种融合用户兴趣表征与注意力机制的推荐算法.利用CVR算法将传统的用户-项目表征转换为用户-兴趣表征;构建一种应用于用户兴趣预测的深度森林模型,引入兴趣簇重要性作为特征选择权重,融合时间注意力机制进行兴趣预测,将用户-兴趣模型结合基于用户的协同过滤算法预测推荐结果.两个数据集上的实验结果表明,该算法能够提高用户兴趣预测准确率,提升推荐效果.     

基于用户动态兴趣的视频点击率预测模型

针对经典的点击预测模型无法捕捉用户动态兴趣和分析特征低阶高阶交互困难的问题,提出一种基于用户动态兴趣的视频点击预测模型。该模型首先将离散的数据经嵌入过程后映射成易于操作的低维连续向量;为捕捉用户动态兴趣变化,引入transformer模型,同时分析用户点击视频序列与待预测的候选视频,抽取行为序列中的视频与待推荐视频之间的相互作用;为深入挖掘用户点击行为背后的隐式特征交互,引入DeepFM网络并在网络结构上进行优化改进,使模型更加适合顺序依赖的点击数据。实验结果表明本文提出改进的模型预测精度优于在点击率预测方面比较具有代表性的深度分解机等模型,同时引入transformer机制可以明显提升点击率预测的精度。     

融合用户偏好和多交互网络的协同过滤算法

推荐系统中的辅助信息可以为推荐提供有用的帮助,而传统的协同过滤算法在计算用户相似度时对辅助信息的利用率低,数据稀疏性大,导致推荐的精度偏低.针对这一问题,本文提出了一种融合用户偏好和多交互网络的协同过滤算法(NIAP-CF).该算法首先根据评分矩阵和项目属性特征矩阵挖掘出用户的项目属性偏好信息,然后使用SBM方法计算用户间的项目属性偏好相似度,并用其改进用户相似度计算公式.在进行评分预测时,构建融合用户-项目属性偏好信息的多交互神经网络预测模型,使用动态权衡参数综合由用户相似度计算出的预测评分和模型的预测评分来进行项目推荐.本文使用MovieLens数据集进行实验验证,实验结果表明改进算法能够提高推荐的精度,降低评分预测的MAE和RMSE值.     

新闻

众筹创业的困扰 作为最成功的众筹平台，Kickster将创意人才与愿意资助他们实现梦想的人联系起来，迄今已有3万个项目通过上百万人的资助实现了融资目标。     

A social recommendation approach for reward-based crowdfunding campaigns

In recent years, a new kind of fundraising mode known as crowdfunding has gradually emerged. Because of the rapid spread of the internet, people can offer their creative ideas on a fundraising platform and attract mass backers to invest in their projects. Crowdfunding not only helps users to realize their dreams but also allows companies to carry out test marketing. Although crowdfunding brings huge opportunities, the success rate of fundraising plans remains low. In this study, we therefore propose a phase-based backer recommendation mechanism, which integrates information from crowdfunding and social networking platforms by analyzing the factors of social relationships, user preferences, and economic backgrounds, to help project creators to reach their fundraising goals at each stage. Our experimental results show that the proposed mechanism is effective in identifying appropriate backers with respect to the status of the project and significantly improves the success rates of crowdfunding. The proposed mechanism can provide greater business value and more opportunities to crowdfunding platforms and contribute to more successful fundraising plans.     

Money is not everything: A typology of crowdfunding project creators

While studies focus on how crowdfunding promotes the launch of innovative products or services through two-sided (creator-investor) platforms and digital tools, knowledge on the creator-side motivation is limited. To address this knowledge gap, our study identifies four types of crowdfunding project creators—social entrepreneur, fund seeker, indie producer, and daring dreamer—based on four motivations—achievement, monetary need, prosociality, and relationship building. We also examine how crowdfunding projects’ characteristics differ by the creator type and which characteristics are critical for enhancing projects’ performance. The study extends crowdfunding literature by deepening knowledge of creator-side motivation and contributes to crowdfunding practices by suggesting platform operators ways to attract and promote different types of project creators.     

A user requirement driven framework for collaborative design knowledge management

In today’s competitive global marketplace and knowledge-based economy, user requirement becomes an important input information for enterprises to develop new product and a critical factor to drive product collaborative design evolution. Meanwhile, there remains no consensus on how best to support knowledge activities and significant challenges to establishing design information management facing to rapid collaborative product development with dynamic user requirement. This paper introduces solutions for designer to deal with dynamic user requirement information through requirement evaluation and prediction method. In this study, we propose a user requirements-oriented knowledge management concept that is based on a four level hierarchy map model with special regard to knowledge collaboration and information communication. Furthermore, a novel distributed concurrent and interactive user requirement database was constructed, and the framework driven by user requirement was put forward to support collaborative design knowledge management. Finally, the service robot design project of a start-up company is used as a case study to explain the implementation of proposed framework.     

基于协同工作的PDM项目开发平台模型

文章借助CSCW的思想对PDM 中的项目协同设计过程进行了分析,结合面向对象的描述方法,给出了项目协同设计系统的静态与动态描述模型。运用分布式技术,给出了一个基于协同工作的PDM项目开发平台模型。     

An investigation of enablers and inhibitors of crowdfunding adoption: Empirical evidence from startups in China

In recent years, crowdfunding has emerged as a new fundraising technique for startup ventures. Crowdfunding allows small businesses and startups to seek funding from a large number of private investors, the so‐called “crowdfunders.” Crowdfunding is beneficial for startup ventures by offering a variety of financial and nonfinancial benefits such as helping to solve funding difficulties, providing value‐added involvement, facilitating access to further funding, and providing publicity. In contrast, diffusion processes of new business innovations have become increasingly complicatedly and multifaceted in recent years. Potential users today are exposed to a wide range of influences that include word‐of‐mouth communications, network externalities, and social signals. Despite the advantages associated with crowdfunding, the crowdfunding industry faces slow rates of diffusion in the Chinese market because crowdfunding ventures often fail to translate into adoption behavior. This study investigates the enablers and inhibitors of crowdfunding adoption intention from the perspective of startups by employing the two‐factor theory, status quo bias theory, and innovation diffusion theory. These theories allow crowdfunding researchers and platform managers to test the relative influence of both reasons for (enablers) and reasons against (inhibitors) crowdfunding adoption in a single framework. In conclusion, considering the phenomenon of startups in China, the results of this study will fill the paucity of empirical research in the innovation resistance area, shedding light on the motivation of entrepreneurs to avoid crowdfunding as a financing course as well as providing crowdfunding managers and policymakers with strategic recommendations.     

How do investors decide? An interdisciplinary review of decision-making in crowdfunding

Crowdfunding is on the rise: its volume grew 1000% in only three years and is about to outpace worldwide venture capital spending. A quickly growing body of research is exploring the emerging crowdfunding phenomenon. While the literature offers a detailed and comprehensive picture of decision-making for traditional startup financing or bank loans, it does not provide a holistic understanding of decisions to invest in crowdfunded ventures. Many individual studies investigate isolated factors that influence investor decision-making in crowdfunding campaigns without integrating the findings regarding those influences. A comprehensive view of the relevant decision-making factors is necessary to build future research on and for practitioners to gain a better understanding of how investors choose. We conduct an interdisciplinary literature review to examine which factors influence the investment decisions in crowdfunding. From an analysis of 68 articles, we construct a comprehensive framework of relevant influence factors. Even though prior research covers many factors, others have received scant attention. Especially investors’ cognitive features and the context in which the investment decision is made seem to strongly influence decisions but are scarcely researched. In addition, most reviewed studies focus more on individual factors and campaign success than underlying decision processes. To highlight novel factors of crowdfunding investment decisions, we compare decision-making in traditional investments, such as venture capital and bank loans, to crowdfunding. Our findings offer new avenues for research toward understanding how the shift induced by crowdfunding changes our choices and actions. The analysis should support the endeavor to build better theories and provide a basis for further social and technological development.     

Dynamics of investor communication in equity crowdfunding

In crowdfunding, start-ups can voluntarily communicate with their investors by posting updates. We investigate whether start-ups strategically use updates, which have previously been shown to increase investments. To this end, we use hand-collected data from 751 updates and 39,036 investment decisions from two major German equity crowdfunding portals: Seedmatch and Companisto. We find evidence of strategic communication behavior of start-ups during an equity crowdfunding campaign. During the funding period, start-ups more frequently post updates with linguistic devices that enhance the group identity and the group cohesion as well as updates on the business development. Furthermore, the probability of an update during the funding period increases along with strong competition of other contemporary crowdfunding campaigns.     

Feature subset selection for predicting the success of crowdfunding project campaigns

Electronic Markets - Statistics from crowdfunding platforms show that a small percent of crowdfunding projects succeed in securing funds. This makes project creators eager to know the probability...     

融合协同过滤和主题模型的弹幕视频推荐算法

针对弹幕视频网站通常不为用户提供评分功能并且使用弹幕池策略以降低存储压力导致的缺少用户评分和无法准确建模用户主题的问题,提出融合主题模型和协同过滤的弹幕视频推荐算法（DRCFT）。首先根据用户行为构造隐式评分矩阵,并得到项目评分相似度;其次建立LDA主体模型,生成项目主题,计算项目主题相似度;再加权生成综合项目相似度,由此结合评分矩阵计算用户—项目预测评分;然后对项目主题相似度矩阵进行sigmoid惩罚,结合用户历史记录得到用户—项目主题相似度;最后将用户—项目主题相似度作为权重,与预测评分相乘,得到最终的预测评分。对比实验表明,该算法能够得到合理的推荐结果,提高推荐的准确性。