Computational of Concrete Slump Model Based on H2O Deep Learning framework and Bagging to reduce Effects of Noise and Overfitting

Abstract

Concrete mixture design for concrete slump test has many characteristics and mostly noisy. Such data will affect prediction of machine learning. This study aims to experiment on H2O Deep Learning framework and Bagging for noisy data and overfitting avoidance to create the Concrete Slump Model. The data consists of cement, blast furnace slag, fly ash, water, superplasticizer, coarse aggregate, fine aggregate, age, slump, and compressive strength. A primary data for concrete mixed design using the fine aggregate material from Merapi Volcano, the hills of Muntilan, and Kalioro. The coarse aggregate was obtained from Pamotan, Jepara, Semarang, Ungaran, and Mojosongo Boyolali Central Java. The cement was using Gresik and Holcim product and the water was from Tembalang, Semarang. The experiment model with one input layer with 7 neurons, one hidden layer with 20 neurons, and one output layer with 1 neuron using activation function TanH, with parameter L1=1.0E-5, L2=0.0, max weight=10.0, epsilon=1.0E-8, rho=0.99, and epoch=800 is able to achieve RMSE of 2.272. This result shows that after introducing Bagging, the error can be reduced up to 2.5 RMSE approximately (50% lower) compared to the model without Bagging. The manually tested mixture data was used to model evaluation. The result shows that the model was able to achieve RMSE 0.568. Following this study, this model can be used for further research such as creating slump design practicum equipment/ application software.

Enhancing Subsurface Soil Moisture Forecasting: A Long Short-Term Memory Network Model Using Weather Data

“…In Figure 10, it is shown that the training and validation loss moves downward in the 51st epoch, and their difference also decreases at that point, and the fluctuation is lower. However, as the graph shows a little noise in the training and validation loss curve in the last six epochs, there can be a chance of being slightly overfitted [70], which was tested later in the model evaluation section. [61] 0.93 11, the high accuracy of the model at the 51st epoch is evidence of no underfitting.…”