Instability in important socioeconomic indicators can have far-reaching effects on global development. This thesis offers a set of one-of-a-kind big data analytics algorithms that operate on unstructured Web data streams to automatically infer events, knowledge graphs, and predictive models, allowing for a better understanding, definition, and anticipation of the volatility of socioeconomic indicators. This paper we presents four major results that expand previous knowledge. Given a large volume of diverse unstructured news streams, we first describe novel models for collecting events and learning spatio-temporal features of events from news streams. We explore two different kinds of event models: one that is based on the concept of event triggers, and another that is probabilistic and learns a generic class of meta-events by extracting named entities from text streams. The second piece of work investigates the challenge of gleaning knowledge graphs from time-sensitive data like news and events as they happen. Event graphs produce a condensed depiction of a chronology of events pertinent to a news query by characterizing linkages between them using "event-phenomenon graphs," while spatio-temporal article graphs capture innate links between news stories. In this paper we present the various result outcome for predictive result analysis.