In this section, we will continue to improve the accuracy of the stock price prediction by integrating historical price points data with the most-recent headlines of the company for which we are predicting the stock price. 

The strategy that we will adopt to integrate data from multiple sources—structured (historical price) data and unstructured (headline) data is as follows:

• We will convert the unstructured text into a structured format in a manner that is similar to the way we categorized news articles into topics.
• We will pass the structured format of text through a neural network and extract the hidden layer output.
• Finally, we pass the hidden layer output to the output layer, where the output layer has one node.
• In a similar manner, we pass the input historical price data through the neural network to extract the hidden layer values, which then get passed to the output layer that has one unit in output.
• We multiply the output of each of the individual neural network operations to extract the final output.
• The squared error value of the final output shall now be minimized.