To investigate the non-linear hydrodynamics in horizontal gas-liquid bubbly flow, time-series data of local void fraction were obtained using a double sensor probe. Both deterministic chaos and stochastic analyses were used to diagnose the dynamic behaviors in the system. The power spectra of the local void fraction exhibited no single distinct peak, indicating the aperiodic nature of the variation of the local void fraction. The kolmogorov entropy and correlation dimension of the attractor reconstructed from the time-series data by the embedding method were to be positive and very large, the Hurst exponent computed from the rescaled range analysis was found to be larger than 0.5, indicating the non-linear chaotic feature of the system. The Kolmogorov entropy and the correlation dimension decrease with increasing liquid superficial velocity, and the correlation dimension changes less if the gas superficial velocity is high. On the other hand, the fluctuation of local void fraction in horizontal gas-liquid bubbly flow could be considered to take place due to the zigzag motion of bubbles with different shape, size and velocity.