Background on Traffic Flow:
Traffic flow is one of the most studied phenomena in urban planning and transportation engineering. As cities grow and the use of vehicles increases, understandng and predicting traffic conditions becomes a critical challenge with direct consequences for public safety, environmental sustainability, and our productivity in general.
Traffic flow is typically monitored through a combination of fixed infrastructure and mobile sensing technologies. Inductive loop detectors embedded in road surfaces measure vehicle counts and speeds at fixed points. Camera-based systems and radar sensors provide additional spatial coverage, while GPS data collected from smartphones and connected vehicles has enabled large-scale, real-time traffic monitoring at a network level. These systems generate continuous streams of multivariate time-series data, including vehicle counts by type, time of day, and day of week, all of which are represented in the dataset used in this module.
Multilayer Perceptron (MLP):
A Multilayer Perceptron is a foundational class of artificial neural network used for both classification and regression tasks. It consists of an input layer, one or more hidden layers of neurons, and an output layer. Each neuron applies a weighted sum of its inputs followed by a nonlinear activation function, typically ReLU in modern networks, which allows the network to learn complex nonlinear decision boundaries that simpler models cannot capture.
Hybrid Model:
In this module, the hybrid model follows a three-stage architecture. A small classical neural network first compresses the input features into a low-dimensional representation sized to match the number of available qubits. This compressed representation is then passed through a quantum layer, a PennyLane quantum circuit with trainable gate parameters that transforms the input through a high-dimensional quantum Hilbert space. Finally, a classical linear output head maps the quantum layer's expectation values to class probabilities.