| 摘要: | 併購是企業中可以達到永續經營的手段。然而之前很少人探討關於預測併購交易是否成功的研究。因此，如果在公司要進行併購交易之前，能夠有一個預測模型，來預測併購後成功與否。不僅可以讓公司的管理者在併購決策上提供幫助，也可以讓投資者做出更明智的投資決策。
本論文結合文本與數值特徵預測併購成敗。從10-K文件中的管理層討論與分析(MD&A)來提取文本特徵，透過分層注意力網路來建構文本向量，並使用公司每三年的MD&A所計算的MD&A時間變化量，以及結合15項財務指標，來當作模型的輸入因子，使用貝氏神經網路進行訓練。
本研究旨在預測併購成功，然而在實驗結果的表現不如預期，可能來自內部因素或外部因素的影響，包括財務狀況、併購策略等原因，因此只使用財務指標或加入文本數據來進行預測，有可能造成預測結果不準確。應當將影響併購的因素更全面化地考慮，並透過特徵選取，提取只對預測併購成功有影響的特徵進行訓練，並在做出最終決策之前，綜合考慮各個模型的結果、相關領域的專業知識以及市場情況等多方面的信息。;Mergers and acquisitions are the means by which enterprises can achieve sustainable operation. However, little research has been done on predicting the success of M&A deals before. Therefore, if the company is going to conduct an M&A transaction, it can have a predictive model to predict whether it will be successful after the M&A transaction. Not only can the company′s managers provide assistance in M&A decisions, but it can also allow investors to make more informed investment decisions.
This paper combines text and numerical features to predict the success or failure of mergers and acquisitions. Extract text features from the management discussion and analysis (MD&A) in the 10-K file, construct the text vector through Hierarchical Attention Network, and use the MD&A time change calculated by the company′s MD&A every three years, and combine 15 financial indicators are used as the input factors of the model, and use Bayesian neural network to train model.
This study aims to predict the success of mergers and acquisitions. However, the performance of the experimental results is not as expected, which may come from internal factors or external factors, including financial conditions, merger strategies and other reasons.
Therefore, only using financial indicators or adding text data to forecast may cause inaccurate forecast results. The factors affecting mergers and acquisitions should be considered more comprehensively, and through feature selection, only the features that have an impact on predicting the success of mergers and acquisitions should be extracted for training. And before making a final decision, comprehensively consider the results of various models, professional knowledge in related fields, and market conditions and other information. |
