Algorithm 1 CNN-LSTM-GA: Feature selection and hyperparameter optimization

Input: D—Dataset, $N_{pop}$—population size, $N_{gen}$—number of generations, $p_c$—crossover probability, $p_m$—mutation probability   Output:  Best individual $I^{*}$ and corresponding CNN-LSTM model $M^{*}$     1:Split D into train (40%), validation (40%) and test (20%) subsets; // Initialize population P with $N_{pop}$ individuals   2:for each individual $i\in P$ do   3:      Randomly initialize feature subset $F_i$ and model hyperparameters $H_i$;   4:end for   5:for generation = 1 to $N_{gen}$ do   6:      for each individual $i\in P$ do   7:            Decode chromosome → feature subset $F_i$ and model hyperparameters $H_i$;   8:            Build CNN-LSTM($H_i$);   9:            Construct $T_{train}$ and $T_{test}$ datasets from EEG data using feature subset $F_i$; 10:            Train CNN-LSTM($H_i$) model with $T_{train}$; 11:            Compute fitness $f\left(i\right)$ as the F1-score of CNN_LSTM($H_i$) evaluated on $T_{validation}$; 12:      end for 13:      Apply tournament selection to choose parents based on $f\left(i\right)$; 14:      For each parent pair, generate two offspring using crossover with probability $p_c$. 15:      for each offspring gene g with probability $p_m$ do 16:            if g is Boolean then 17:                  Flip gene g; 18:            else if g is Integer or Real then 19:                  Apply window mutation; 20:            else if g is Nominal then 21:                  Replace with random category; 22:            end if 23:      end for 24:      Form new population P combining best individuals and offspring; 25:end for 26:Select best individual $I^{*}=arg{max}_{i}f\left(i\right)$; 27:Decode individual $I^{*}$→ feature subset $F_i$ and model hyperparameters $H_i$; 28:Build CNN-LSTM($H_i$); 29:Construct $T_{train}$ and $T_{test}$ datasets from EEG data using feature subset $F_i$; 30:Train CNN-LSTM($H_i$) model with $T_{train}$; 31:Evaluate final model $M^{*}$ on $T_{test}$; 32:return $I^{*},M^{*}$