A reliable prediction of net radiation (Rn) inside naturally ventilated greenhouses is critical for accurate evapotranspiration evaluation and thus for water saving, considering that previous studies have indicated that evapotranspiration in such relatively decoupled greenhouses is predominantly controlled by greenhouse Rn (Rn-GH). We hypothesized here that Rn-GH in naturally ventilated greenhouses can be accurately predicted using global solar radiation in the vicinity of the greenhouse (Rs-out) as the only measured parameter, together with the calculated position of the sun, defined by the solar elevation angle and solar azimuth. To test this hypothesis, we performed experiments in two adjacent greenhouses in the Southern Negev, Israel (30.96° N, 34.69° E) under arid climate. In one of the greenhouses, tomato was grown during winter 2017–2018, while in the other, melon was grown during winter and spring 2018–2019. Our analyses demonstrated that Rn-GH can be accurately predicted (r2 = 0.982) using Rs-out as the only measured parameter, while the global solar radiation inside the greenhouse (Rs-GH), and the ratio between Rn-GH and Rs-GH are predominantly dependent on solar elevation angle and solar azimuth, as well as the greenhouse structure and cloud cover. This paper shows that the impact of these properties on the association between Rs-out and Rn-GH can be accurately resolved using multivariate regression by the k-nearest neighbors approach. This suggests that computerized modeling of the greenhouse structure and light transmission can potentially enable precise evaluation of Rn-GH and therefore also reference evapotranspiration in naturally ventilated greenhouses, using Rs-out as the only measured parameter. A calculation-based factor for the cloud effect on Rs-out transmittance into the greenhouse significantly improved the Rn-GH prediction under cloudy conditions.