Recently, cutting load analysis methods are used for tool wear/breakage diagnosis, equipment maintenance and machining data optimisation. In addition, by measuring the cutting load generated during machining, it is possible to verify machining conditions according to equipment, tool and material combinations, and to contribute to productivity and quality improvement. High-speed sampling (10 kHz or higher) using acceleration or current sensors can be used to predict not only tool condition but also machined surface roughness. This study addresses how machining load, which has been used to diagnose tool wear and spindle condition, can be used to measure the quality of machined surfaces. This study presents a method for extracting patterns from sensor data values in the machining history data, predicting machined surface roughness. The surface roughness value was predicted by the regression equation derived from the mean of the machining load and the sum of the mean machining load and the magnitude of the dominant frequency. The average error between the proposed method and the actual surface roughness value was 0.172 μm.