This paper studied the performance analysis of the chromatic dispersion problem in optical networks using deep learning approaches. An end-to-end autoencoder is an essential estimation parameter for replacing the transmitter and receiver in the optical fiber communication system's intensity modulation direct detection (IM-DD). We need a generative adversarial network (GAN) and square-law detector to estimate gradient return problems in end-to-end optical fiber transmission and reduce the complexity of the networks. The major estimation parameters in optical networks are the bit error rate and signal-to-noise ratio. The simulation results represent the bit error rate and signal-to-noise ratio of the prospective technique i.e., crucially varies compared to the optical and electrical domain algorithms. We need to improve performance with small amounts of dispersion; a windowed design using a frequency domain description of chromatic dispersion may be used. This work discusses a dispersion compensation technique to reduce the chromatic dispersion and strengthen system performance. The simulation results are carried out by using MATLAB.