Quantitative models are changing virtually every aspect of investment. In this thesis, we focus on the application of machine learning and financial network in investment. On the one hand, machine learning models can be used to detect complex patterns among financial data and make predictions about the market in the future. On the other hand, network science and topology facilitate the understanding of the structure that governs a complex system. Given the intricate and hierarchical nature of the financial market, it is vital to develop new network models for a better comprehension of its mechanism. The rest of the thesis is organized as follows. In the first chapter, we construct a fi- nancial network among portfolios based on their common asset holdings and propose a cascade mechanism to explain how the linkage in the financial network can influence the portfolio returns. Then we apply the network structure in the design of a regularization function for a vector autoregression model, with the purpose to predict portfolio returns. In the second chapter, we devise a strategy to exploit arbitrage opportunities due to cas- cade behaviors among investors. The behaviors are detected with structural break tests while moving average methods are used to predict market directions. We then apply machine learning methods to intensify the strategy. Additionally, volatility prediction for stock return is another important topic in quantitative investment, as it is essential in vari- ous fields such as option pricing, portfolio allocation, and risk management of portfolios. In the third chapter, we propose a machine learning-based method for daily volatility prediction which outperforms existing methods in terms of various prediction errors.