Building a Crypto Price Predictor Using Google Cloud Platform - Part 1 - Committing Myself to Working on This

Why am I Building a Crypto Price Predictor?

First things first: I AM NOT A FINANCIAL ADVISOR.

I'm not trained in this stuff, and I do not plan to try to convince you to invest in crypto-currencies. If you asked for my investing advice, I'd tell you to dump your savings into an all-market ETF, buy a small percentage of fixed return bonds, and set it and forget it.

I am a BogleHead.

This blog post was inspired by recent discussions with my boss.

He is convinced that Crypto will enter a strong bull market soon and could be a good windfall for making some money💰 (HE ALSO IS NOT A FINANCIAL ADVISOR).

After years of neglect, I decided to start looking into crypto again, and the first thing I noticed was the sweet, sweet variance.

Statistics for Digital Marketers: Descriptive Statistics Crash Course

Go beyond the default insights offered by GA4.

DataWithJavi.comJavier Cabrera

I thought I could dedicate a small percentage of my portfolio to trading through this variance while working on my data skills.

Specifically, building a data product leveraging Google Cloud Platform (GCP).

• The Idea: Build a price predictor tool to help me make winning trades in my crypto day trading journey.
• The probable outcome: I'll lose some money but learn a ton in the process.

Why am I posting this before I've built the product?

I want to commit to building something, so by publishing this, I hope that you, dear reader, will hold me accountable if I don't.

So, What's the Plan?

As mentioned before, I want this project to be an opportunity to practice and develop my GCP skills.

I have started the work by planning a rough outline of what I want to do in the full scope of this project, although I will probably be doing it in parts. The MVP might be a lot less ambitious than what I present to you today.

We will probably focus on only one or two cryptocurrencies and leverage as few data sources as possible.

My idea for the back-end architecture is the following:

Data Sources

• CoinGecko:
  • API for cryptocurrency prices.
  • It probably will fetch prices daily.
  • Cryptos to be monitored are yet to be defined.
• Reddit API:
  • API to retrieve data from Reddit with the hope of doing a bit of social listening and sentiment analysis.
  • I will focus on crypto-currency subreddits, which are yet to be defined.
• Google Search Trends:
  • Get insights into the search volume of key terms related to cryptocurrencies.
• Other Data Sources:
  • If all goes well, we may input external data into our predictor model, such as US interest rates, SP500 performance, and other signals.

Data Retrieval and Transformation

We will leverage two GCP services for the data retrieval and transformation stage. We don't plan on ingesting a considerable volume of Data with higher-scale projects. You might think of using different ones, but I believe that for our purpose, the following will work:

• Cloud Functions: We will use cloud functions written in Python to retrieve data from the APIs.
• BigQuery: We will push that data to bigquery for storage and modeling.
• Predicting: I'm not sure what we will do for predictions yet. We might build a model using BQML or upload a model built locally to the cloud. This is yet to be defined. We will not use VERTEX AI as the cost can be substantial.

Back-end API and Front-end Website

Honestly, this is the part I am most nervous about, as I have the least experience with it.

For the API, we will probably use Python and a framework such as FLASK to build the distinct endpoints our application will use to retrieve data.

For the front end, I literally do not know. It will be a great time to learn some React, I suppose.

Final Thoughts

I'm not sure about the timeline of this side project. It depends on too many variables. I will tell you that I'm committed to publishing a series of blog posts as I struggle to build this.

If you'd like to find out when the next post comes out, don't hesitate to subscribe 😉