Deep Learning - Recurrent Neural Networks with TensorFlow - Stock Return Predictions Using LSTMs (Part 2)

The video tutorial explains the preference for predicting stock returns over stock prices, using a formula similar to calculating discounts. It details data processing with Pandas, including shifting and subtracting closing prices to calculate returns. The tutorial covers normalizing returns with a standard scaler and training an RNN model, highlighting challenges like overfitting. It concludes with model evaluation through one-step and multi-step forecasts, noting the model's limitations in learning effectively.

Why is it more conventional to predict stock returns rather than stock prices?

• Returns provide a more stable measure over time.

• Stock prices are too volatile to predict.

• Returns are easier to calculate.

• Stock prices are not available for all companies.

What is the formula for calculating stock return?

• Initial price minus final price divided by final price

• Final price divided by initial price

• Final price minus initial price divided by initial price

• Initial price divided by final price

What is the purpose of shifting the closing price in the data frame?

• To align today's closing price with tomorrow's

• To align yesterday's closing price with today's

• To remove missing values

• To calculate the average price

How does the RNN model handle missing data in the first row?

• It fills the missing data with zeros.

• It uses the average of the dataset.

• It ignores the first row during training.

• It predicts the missing data.

Why is data normalization important in the context of stock returns?

• To increase the size of the dataset

• To simplify the calculation of returns

• To make the data more uniform for model training

• To remove outliers from the dataset

What challenge does the RNN model face during training with stock return data?

• The model learns too quickly.

• The model overfits to the noise.

• The model underfits the data.

• The model cannot process the data.

What is the result of the multi-step forecast using the RNN model?

• The model accurately predicts future values.

• The model fails to predict and repeats the same value.

• The model predicts random values.

• The model improves over time.