In this research, the color yield (K/S and Fk) of 100% cotton fabric which is dyed with 6 selected direct dyes is modeled as a function of dye concentration, electrolyte concentration, time and temperature of dyeing. Direct dyes also called substantive dyes can be applied to cotton fibers easily. These dyes are less expensive than the other dyes, suitable for cellulosic fibers, however their washing fastness is a mayor draw back. In order to determine the color yield (Fk), the reflection of samples were measured and then their K/S was calculated. The selected direct dyes were based on diazo group and they were chosen from A, B and C classes. To conduct the experiments using Taguchi approach, two levels were chosen for each factor. After gathering data related to K/S and Fk, significant factors were determined using ANOVA. In the next step, the levels of significant factors in Taguchi approach were increased to 3 and the supplementary experiments were carried out using full factorial design. Having specified significant factors in full factorial design, response surface regression method considering these significant factors was applied in order to model Fk . Then the validity of initial model was evaluated using residual graphs. After verifying the validity of initial models, Box-Cox transformation was implemented until the models enjoyed validity. After the implementation of recommended transformation on Fk (response), using ANOVA leading significant factors were specified once more and modeling process was carried out again in order to achieved a new modified model. The validity of modified model was investigated similar to the initial one. For both initial and modified model, some indicators such as R2, Radj2 and MSE were studied. After investigating these for both initial and modified model, the final model was presented. The results show that dyeing process of 100% cotton fabrics with 6 selected direct dyes can be modeled using response surface regression method. The obtained value of R2 and R2adj of models prove that they possess appropriate fitness in all cases. It has been also shown that there isnt any especial similarity not only among models related to dyes of A, B and C classes but also between pairs of dyes in each group. In the majority of cases, dye concentration, electrolyte concentration and dyeing temprature are the most effective factors on color yield (Fk), respectively. It should be noted that time of dyeing, in the selected range in this research, doesnt affect color yield and doesnt appear in models. Comparison between the results of these two methods, Taguchi and full factorial designs, and central composite desig show that descriptive indicators (R2 and adj2) of the obtained models using Taguchi and full factorial designs, has been impoved for all selected direct dyes.