Predictions through an artificial neural network

Artificial neural networks are mathematical and computational models built upon the theory of how biological neurons in the brain work. They are able to adapt rules provided by the user and can learn new ones through a feed of training data provided to them.

Researchers from Oman, Kuwait and Syria performed three sets of experiments using a type of backpropagation neural network (BPNN). They tried to predict humic-substance agglomeration and coagulation with heavy metals and polyelectrolytes. Anionic humic substances are generally used to coagulate smaller substances for their removal from liquids. The study compared results obtained from the BPNN tests with previous experimental results to see if they were accurate.

The paper suggests that BPNNs can be used to predict the coagulation and agglomeration of humic substances with an accuracy of >95%. The researchers point out that the accuracy is strongly related to having an optimum number of data points and neurons. A lower than optimum number of either leads to lower accuracy due to a poor learning process. A higher than optimum number is also unreliable because the neurons memorize the input data rather than learning from them.

Sign-up to receive our e-alert update every two weeks to keep up with everything new on the portal.