The stock market is equivocal and unforeseeable because it heavily depends on several macros and micro factors of digital economical market. The aim of this work to solve the early classification problem for accurate prediction of the stock market using Two-Layered Bidirectional Long Short Term Memory (TL-BLSTM) with transfer learning. The TL-BLSTM model is pre-trained with US-based S&P 500 index data and further trained on the National Stock Exchange (NSE) of India such as India Tobacco Company Ltd. Oil and Natural Gas Corporation Ltd. (ITC), Oil and Natural Gas Corporation Ltd. (ONGC), and State Bank of India (SBI). The performance of the TL-BLSTM model is tested and evaluated through three different training strategies using variants of Recurrent Neural Network (RNN) architectures based on Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU). The proposed TL-BLSTM model considerably outperforms than existing RNN based models like LSTM, GRU, and Bidirectional architectures of a hybrid model with LSTM and GRU.