We developed an eruption forecasting model using data from multiple sensors or data streams with the Bayesian Network method. The model generates probabilistic forecasts that are interpretable and resilient against sensor outage. We applied the model at Whakaari/White Island, an andesite island volcano off the coast of New Zealand, using seismic tremor recordings, earthquake rate, and CO2, SO2, and H2S emission rates. At Whakaari/White Island, our model shows increases in eruption probability months to weeks prior to the three explosive eruptions that were recorded since 2012. We also observe that explosive eruptions consistently occurred after the cumulative probability exceeded at least 80%. Although developed for Whakaari/White Island, our model can be easily adapted to other volcanoes, complementing existing forecasting methods that rely on single data streams.