π§© Time Series Puzzle: Given the prices of the stock in the past, predict its value in the future

π  Start Solving

π Explore Dataset

## π΅πΌββοΈ What is The Timeseries Prediction Puzzle About?

Ability to predict the future can be really valuable. But since we donβt have Doc and his DeLorean time machine from Back to the Future, we have to use data science to predict the future.

In this puzzle, you will predict the value of a stock in the future.

## πͺπΌ What Youβll Learn

In this puzzle, you will learn

1. How to predict and forecast the future price of any series
2. How to use the most famous Prophet method developed by Facebook

Letβs get started! π

Your task is to use time-series prediction to find the future value of these synthetic stock prices. Given the prices of the stock in the past, predict its value in the future.

The dataset contains stock prices from 1985-01-29 to 2010-03-25 in train.csv and 2010-03-26 to 2013-06-21 in val.csv set, leaving out the weekend, i.e., Saturday and Sunday. You need to predict the stock prices from 2013-06-24 to 2021-01-13 on the weekdays.

The dataset contains the Training set and Validation set:

1. Training Set: 6345 samples
2. Validation Set: 817 samples

Training & Validation Set The training and validation set contains the dates and values in CSV format. The CSV file contains two columns.

date value
1985-01-29 1552.519959
1985-01-30 1576.070007
1985-01-31 1561.440002
1985-02-01 1554.839935

1. date - date for the stock market
2. value - the value of the stock on the given date

## π Dataset Files

The following files are available in the resources section:

• train.csv - (6345 samples) This training CSV file contains the date and the values.
• val.csv - (817 samples) This validation CSV file contains the date and the values.
• sample_submission.csv - (1905 samples) File used to evaluate the leaderboard score but does not have the values.

## π¬ Let's Solve This Puzzle

Click here to access the basic starter kit. It contains in-depth instructions to:

2. Setup the AIcrowd-CLI environment that will help you submit directly via a notebook
4. Preprocessing the dataset
5. Creating the model
6. Setting the model
7. Training the model
8. Submitting the result

Make your first submission using the starter kit. π

## π Evaluation Criteria

The Mean Squared Error Metric is used here to test the efficiency of your model.

## π€« Hint to get started

Check out the open-source library Prophet, developed by Facebook and designed to automatically forecast univariate time series data. You can learn more about it here.

## π Resource Circle

Learn about another state-of-the-art approach using LSTM here.