We present a new pipeline to boost the computational efficiency and generalization of imbalanced time series analysis. Key components of the pipeline are – (i) a self-reliant topological summary estimation of time series and (ii) a topological augmentation method to generate a balanced analysis. By using sublevel set filtration, our proposed pipeline constructs topological summaries that rely only on the local extrema of the time series and circumvent the necessity of delay embedding of the time series. Such topological summary estimation combined with the topological augmentation method delivers a powerful mechanism to explore imbalanced time series data. This mechanism can be deployed for a variety of statistical analyses and machine learning tools. The effectiveness of the pipeline is demonstrated under two scenarios: one involving bankruptcy prediction of the United States (US) companies between 2000 and 2023, and the other involving hypothesis tests to distinguish between US stock indices obtained with recession and no recession periods between 2000 and 2023.