High performance computing (HPC) capabilities equate directly to competitive advantage in financial sector. The priority in parallelism enables significant improvements of cash flow optimization with high frequent computing, satisfying the ever-increasing demand for real-time analysis and processing of large dataset.The main goal of this thesis is to derive a combined approach of Hybrid Parallel Genetic Algorithm (HPGA) and N-gram Stochastic Language Model, to facilitate the optimal financing and investment issue of emerging small business. Major works are as listed below,1. Briefly review the history and applications of HPGA, HPC, and language modeling, subsequently analyze their theories and characteristics.2. Apply N-gram model to extract text properties of enterprise annual reports, realize automatic evaluation of risk sentiment with language model features and readability calculations, and make predictions of performance and credit ranking by SVM classifier and event study.3. Build statistical models for high frequent cash flow optimization with results from annual report predictions, to fulfill the requirement of GA framework and search space compression.4. Design a new HPGA algorithm based on In-Memory Database to overcome shortcomings, and implement the HPGA on HPC and In-Memory computing platform, complete the cash flow optimizing with higher speed and efficiency. |