A Hybrid Personalized Recommender System for Content Suggestion

拟议的研究是在推荐系统和预测分析领域，并应在加拿大ICT（信息和通信技术）部门的影响。PostBeyond是一家总部位于多伦多的公司，主要为大型企业提供员工通信解决方案。使用PostBeyond的产品，公司员工可以通过一个中央内容中心轻松访问公司相关信息，并可以选择通过他们的社交网络与外部分享信息。目前的解决方案是一个尺寸适合所有。有时，员工很难找到他们特别感兴趣并想与他人分享的内容。为了解决这个问题，在这个项目中，我们提出了建立一个个性化的推荐系统，可以建议内容的员工访问，消费和共享的基础上，他们的个人喜好。我们必须解决的几个研究挑战包括：1）缺乏关于用户内容访问和共享模式的历史数据; 2）需要定义通用和可扩展的内容配置文件; 3）需要将用户的动态行为纳入系统; 4）难以有效地将群组推荐功能集成到个性化推荐系统中。为了成功构建推荐系统，本文确定了以下三个任务：1）开发日志工具以保存用户的内容访问和社交分享记录; 2）应用日志挖掘技术来挖掘用户的模式、行为，并观察一般的内容消费或分享趋势; 3）设计并实现一个混合推荐系统来进行个性化的内容推荐和预测。拟议的项目可以使PostBeyond受益，使他们目前的通信解决方案更有效，并针对每个员工的行为进行定制，从而对他们的潜在客户更具吸引力。它还为PostBeyond和许多使用其解决方案的加拿大企业带来了经济效益。

The proposed research is in the area of recommender systems and predictive analytics, and should have animpact in Canadian ICT (Information and Communications Technology) sector. PostBeyond is aToronto-based company providing employee communications solutions to primarily large enterprises. UsingPostBeyond's product, company employees can have easy access to company-related information through acentral content hub, with the option to share the information externally across their social networks. Thecurrent solution is one size fit all. Sometimes it could be hard for individual employees to find content they areparticularly interested in and would like to share with others. To solve this problem, in this project, we proposeto build a personalized recommender system that could suggest content for employees to access, consume andshare based on their individual preferences. A few research challenges we must tackle include: 1) the lack ofhistorical data on users' content access and sharing patterns; 2) the need to define content profile that is bothgeneric and extensible; 3) the need to accommodate users' dynamic behavior into the system; 4) the difficultyof effectively integrating group recommendation features into the personalized recommendation system. Tosuccessfully build the proposed recommender system, the following three tasks are identified: 1) developing alogging tool to save users' content access and social sharing records; 2) applying log mining techniques to minethe user patterns, behaviors and observe the general content consumption or sharing trends; 3) designing andimplementing a hybrid recommender system to do personalized content suggestion and prediction. Theproposed project could benefit PostBeyond by making their current communications solutions more efficientand customized towards each individual employee's behavior, and thus a more attractive solution to theirpotential customers. It also drives economic benefits for both PostBeyond and the many Canadian enterprisesthat use their solutions.