Real-time incident detection on freeways plays an important part in any modern traffic management operation by maximizing road system performance. The US Department of Transportation (US-DOT) estimates that over half of all congestion events are caused by highway incidents rather than by rush-hour traffic in big cities. An effective incident detection and management operation cannot prevent incidents, however, it can diminish the impacts of non-recurring congestion problems. The main purpose of real-time incident detection is to reduce delay and the number of secondary accidents, and to improve safety and travel information during unusual traffic conditions. The majority of automatic incident detection algorithms are focused on identifying traffic incident patterns but do not adequately investigate possible similarities in patterns observed under incident-free conditions. When traffic demand exceeds road capacity, density exceeds critical values and traffic speed decreases, the traffic flow process enters a highly unstable regime, often referred to as “stop-and-go” conditions. The most challenging part of real-time incident detection is the recognition of traffic pattern changes when incidents happen during stop-and-go conditions. Recently, short-term freeway congestion detection algorithms have been proposed as solutions to real-time incident detection, using procedures known as dynamic time warping (DTW) and the support vector machine (SVM). Some studies have shown these procedures to produce higher detection rates than Artificial Intelligence (AI) algorithms with lower false alarm rates. These proposed methods combine data mining and time series classification techniques. Such methods comprise interdisciplinary efforts, with the confluence of a set of disciplines, including statistics, machine learning, Artificial Intelligence, and information science. A literature review of the methodology and application of these two models will be presented in the following chapters. SVM, Naïve Bayes (NB), and Random Forest classifier models incorporating temporal data and an ensemble technique, when compared with the original SVM model, achieve improved detection rates by optimizing the parameter thresholds. The main purpose of this dissertation is to examine the most robust algorithms (DTW, SVM, Naïve Bayes, Decision Tree, SVM Ensemble) and to develop a generalized automatic incident detection algorithm characterized by high detection rates and low false alarm rates during peak hours. In this dissertation, the transferability of the developed incident detection model was tested using the Dallas and Miami field datasets. Even though the primary service of urban traffic control centers includes detecting incidents and facilitating incident clearance, estimating freeway incident durations remains a significant incident management challenge for traffic operations centers. As a next step this study examines the effect of V/C (volume/capacity) ratio, level of service (LOS), weather condition, detection mode, number of involved lanes, and incident type on the time duration of traffic incidents. Results of this effort can benefit traffic control centers improving the accuracy of estimated incident duration, thereby improving the authenticity of traveler guidance information.