Extraction of Rules for Tuberculosis Diagnosis Using an Artificial Neural Network

Abstract:

The treatment of tuberculosis (TB) is a major challenge throughout the world. The Western Cape Region of South Africa has the highest occurrence of TB in the world. Here, TB is increasing due to improperly managed treatment programmes and inadequate facilities. The development of rules to aid medical practitioners in the early and accurate diagnosis of tuberculosis should prove worthwhile. A method to extract such diagnostic rules from an artificial neural network is presented. These rules accurately represent the knowledge embedded in the “raw” TB data.

Keywords:

Knowledge Processing - Diagnosis in Medicine - Neural Networks - Data Interpretation

• References

• 1 Shavlik JW. Combining Symbolic and Neural Learning. Machine Learning 1994; 12: 321-31.
• 2 Towell GG, Shavlik JW. Refining Symbolic Knowledge Using Neural Networks. Machine Learning 1994; 12: 321-31.
• 3 Viktor HL, Engelbrecht AP, Cloete I. Reduction of Symbolic Rules from Artificial Neural Networks Using Sensitivity Analysis. IEEE ICNN; 1995. Perth, Australia.:
• 4 Viktor HL, Cloete I. Extracting DNF Rules from Neural Networks. International Workshop on Applications of Neural Networks; Torremolinos, Spain: June 1995
• 5 The South African Tuberculosis Control Program: Practical Guidelines. Department of Health; South Africa: 1996