The importance of studying the efficiency of the Iranian capital market has led the present study to pursue the aim of empirical analysis of comparative market theory in the Iranian capital market. Therefore, in this research, an attempt has been made to explore this goal by analyzing irregular behaviors, inspired by the onion model of Sanders' research and focusing on transactions made in the market and index fluctuations. The present study is descriptive-analytical and applied from the perspective of the objective. In the research strategy layer, the archival and quantitative analysis strategy has been selected. The selection layer includes a quantitative single-method method. Finally, the central layer, techniques and procedures, includes the combination of a multilayer perceptron neural network with a water cycle algorithm for parameter optimization in Python software. The data included variables such as closing price, logarithmic return, moving averages, volatility, RSI, trading volume and calendar effects. The accuracy of the models was measured with the Rock curve and the significant differences in the results were evaluated with the Wilcoxon test. The findings show that the level of compliance of the Iranian capital market behavior with the adaptive market hypothesis is compared to the traditional efficient market hypothesis, because the tests of return independence (such as autocorrelation close to zero) support the efficient market hypothesis, but the performance of the MLP-WCA model with an accuracy of 78% and AUC=0.825, along with weak patterns in RSI and momentum, indicates a better explanation of the return behavior by the traditional market hypothesis. The results also indicate a significant impact of investor behavior on changes in stock value and returns. The results show that the buy and hold strategy with a cumulative return of 2.665 has performed better than the simple moving average rule with a return of 1.475 and the combined version of the moving average with volume with a return of 0.282.