FIN313: Suppose you are a data scientist in IMDB and would like to develop a machine-learning model to do the sentiment: Machine Learning and AI for FinTech Assignment, SUSS, Singapore

As the input data may have different lengths, they should be pre-processed using the one-hot encoding so that it can be easily processed by the program. The following is the suggested code:

Stuck with a lot of homework assignments and feeling stressed ? Take professional academic assistance & Get 100% Plagiarism free papers

Get A Free Quote

(a) (Python code) Suppose you would like to use a neural network method to build the machine learning model. The designed neural network architecture is summarised as follows:
• Input layer size: 5,000
• First hidden layer size: 32, followed by a ‘relu’ activation function.
• Second hidden layer size: 16, followed by a ‘relu’ activation function.
• Third hidden layer size: 8, followed by a ‘relu’ activation function.
• Output layer size: 1, followed by a ‘sigmoid’ activation function.
• Optimizer: ‘rmsprop’, loss: ‘binary_crossentropy’, metrics: ‘accuracy’
• Batch size: 512, epochs=10
Use the keras library to implement the above neural network.

(b) Report the training losses, testing losses, training accuracy, and testing accuracy of the above model (results are rounded to three decimal points). Discuss your findings about the results. 

(c) (Python code) Find the optimal value of “epochs” in your neural network model. [Remark: You are encouraged to use plots to improve the clarity of your explanation.] 

(d) (Python code) Propose at least one way to improve your model’s performance. You need to report the accuracy of your improved model on the testing dataset, take a screenshot of your code, and discuss how the chosen method can improve the model’s performance.

(e) Suppose a confusion matrix with 100 samples is used to evaluate the performance of the machine learning model. The results are summarized as follows

(f) (Python code) You realize that the logistic regression model is also commonly used for a classification task. Implement the coding of the logistic regression model by using sklearn.linear_model library, report its testing accuracy and compare the suitability of logistic regression and the above neural network for this task.