Abstract

Traffic flow prediction is a basic aspect to be considered in transportation management and modeling. Attaining precise information on near and current traffic flows has an extensive range of appliances and it further aids in managing the congestion. Numerous conventional models failed at offering precise prediction results due to “shallow in architecture and hand engineered in features”. Moreover, the raw traffic flow information contains noise that might lead to the worst prediction results. Therefore, this paper intends to design an enhanced prediction model on traffic flow using Optimized Deep Convolutional Neural Network (DCNN). The input features or the technical indicators subjected to the optimized CNN are Average True Range (ATR), Exponential Moving Average (EMA), Relative Strength Indicator (RSI) and Rate of Change (ROC), respectively. Moreover, for precise prediction, the weights of DCNN are optimally tuned using a new Improved Lion Algorithm (LA) termed as Lion with New Territorial Takeover Update (LN-TU) model. In the end, the betterment of implemented work is compared and proved over the conventional models in terms of error analysis and prediction analysis.

摘要

交通流预测是交通管理和建模中需要考虑的一个基本方面。获得有关附近和当前交通流的精确信息具有广泛的设备，它进一步有助于管理拥塞。由于“架构浅，特征手工设计”，许多传统模型无法提供精确的预测结果。此外，原始交通流量信息包含可能导致最差预测结果的噪声。因此，本文打算使用优化的深度卷积神经网络（DCNN）设计一种增强的交通流预测模型。经优化 CNN 的输入特征或技术指标分别为平均真实范围（ATR）、指数移动平均线（EMA）、相对强度指标（RSI）和变化率（ROC）。此外，为了进行精确预测，DCNN 的权重使用一种新的改进的 Lion 算法 (LA) 进行了优化调整，该算法被称为具有新领土接管更新 (LN-TU) 模型的 Lion。最后，在误差分析和预测分析方面，比较并证明了与传统模型相比实施工作的改进。