In real-world settings, firms are often connected to each other in the form of a business network. Typical relationships among firms include competition, collaboration, supply-chain, financing, and so on. The structural embeddedness theory posits that a firm’s embeddedness in a business network may influence its competitive business performance. This highlights the theoretical and practical values toward business network mining and analysis. Unfortunately, unlike individual social networks such as Facebook where a large number of user communities are formed and the networking information is made publicly available, information about business networks is not widely available for academic research or commercial applications. This partly explains why numerous research has been devoted to social network analysis but relatively little work is done for business network mining and analysis. Given the huge potential benefits for firms to mine and analyze business networks but yet the reality of lacking effective tools (i.e., design artifacts) for firms to carry out such tasks, the primary motivation of my PhD research is to fill such a research gap by designing a novel framework for automated business network mining and analysis. Given the fact that latent business relationships often exist among firms and these relationships continuously evolve over time, a manual approach for the discovery and analysis of business network is ineffective. Accordingly, the specific technical research problem of my study is to design an effective methodology to mine dynamic and evolving business networks based on large volume of user-contributed textual contents readily extracted from the World Wide Web (Web). The main contributions of my PhD research are summarized as follows: 1) I designed a novel, weakly supervised latent concept learning framework for dynamic business network mining; 2) to analyze firms’ performance (e.g., stock performance) with respect to a business network, I designed a novel business network inference model named Energy Cascading Model (ECM); the ECM model can infer a focal firm’s performance by taking into account the dynamic influence propagated from the related firms that are connected via a business network; 3) empowered by the automatically mined business networks, some econometrics models were developed to facilitate several empirical studies that examined the impact a firm’s structural properties pertaining to a network on its business performance (e.g., unexpected earnings and innovation performance). The primary research methods that drive my research work include the design science method and the econometric analysis method. More specifically, the design science methodology is used to guide the design of the novel business network mining and analysis framework that are underpinned by techniques such as dynamic topic modeling, semi-supervised text mining, probabilistic language modeling, and influence cascading model. On the other hand, the econometric analysis methodology is applied to conduct several empirical studies empowered by dynamic business network mining. Advanced statistical techniques such as Hierarchical Bayesian Analysis (HBA) and finite mixture modeling are exploited to design the relevant econometric analysis models. For evaluating the design of the proposed business network mining method, controlled experiments were carried out based on data related to the firms included in the Forbes 2000 list. Moreover, textual corpora were retrieved from Reuters, Yahoo! Finance, and Twitter. Our experimental results confirm that the proposed business network mining method significantly outperforms other baseline mining methods such as Support Vector Machine (SVM), Conditional Random Field (CRF), Artificial Neural Network (ANN) and so on. By taking business relationships into consideration, the proposed network inference model (ECM) also significantly outperforms other baseline inference models for mid-term directional stock performance prediction. Finally, our empirical studies underpinned by several econometric models also reveal that the structural properties of a firm extracted from a business network (e.g., the betweenness centrality, number of competitors, number of collaborators) have significant influence on its business performance, such as innovation performance and unexpected earning. On one hand, the empirical studies highlight the necessity and the practical value of the business network mining method. On the other hand, our empirical studies also make some theoretical contributions and the findings lead to several practical implications. Although some design artifacts have been developed and some interesting findings are identified through my empirical studies, some aspects of my research work can be improved and extended in the future. For instance, the relationships among firms may not simply be competition or collaboration but depends on specific product contexts or application domains. Accordingly, my future work will aim to design a more fine-grained business network mining method. Moreover, a more sophisticated network inference model that can take into account an array of internal and external factors of firms will be examined. Finally, more empirical studies will be performed to examine the impact of the structural properties of firms on their business performance under a variety of business settings. Keywords: Automatic Business Network Mining; Topic Modeling; Directional Stock Price Prediction; Innovation Performance; Hierarchical Bayesian Model; Finite Mixture Model