OHLCV, Social and Blockchain Cryptocurrency Data Analysis and Price Forecasting

Cryptocurrency Market today counts a market capitalization of $207 Billion, with more than 3000 coins and where main dominant Cryptos, as BTC, ETH, LTC have reached a clear popularity on Social Networks such Twitter, Facebook, Reddit and GitHub. Hence, it is today an important financial reality that attracts a lot of risk lovers and digital coin users. Nevertheless, the ambiguity of Market Nature and the huge volatility makes this market complex and approach to Cryptocurrency analysis a stiff process. It looks distant from exchange market, which appears stable and with low volatility level, and appears more similar with stock. Both in fact, present high degree of risk, but Crypto market results more fragile. All this makes price forecasting an interesting and complex game. Looking at the actual State of Art, the most interesting trend is the application of several machine learning algorithms, such as simple and multiple Linear Regression, Support Vector machine (SVM), Multilayer Perceptron (MLP) to OHCLV financial data. But the lack of seasonality and the continuous volatility drastically afflict models accuracy. Throughout the recent years, Sentiment Analysis has been involved into the Cryptocurrency price forecasting. It is a tool, based on Opinion Mining and Natural Processing Language that allows extracting polarity from Social Posts and Text, a good proxy of investor Sate of Confidence about Market. Most of works consider just Twitter sentiment and Google Trend with daily data sampling frequency. Today, few papers have inferred on Blockchain quantitative features as possible Price spread explanatory variables. Blockchain is the most underlying cryptocurrency technology and it is definable as a distributed, immutable and transparent ledger that allows emitting transactions stored by blocks. This innovation paradigm is impacting on several business areas, as Financial Transactions, Supply Chain and Politic, with a hype expectation that is touching the stars. The scope of this work, is to explore the main Cryptocurrency Sources, and evaluating which kind of data is offered, with which granularity and time horizon and in which ways (REST APIs, Web Socket APIs, csv, excel adds-on). Under this perspective, three kinds of data are stored: the OHLCV (Open, High, Low, Close, Volume) financial data, Social Data, including Facebook likes, Reddit posts, comments, GitHub activity and Blockchian data, as Block size in Byte, the number of Transactions, the Difficulty to add a new Block, the Miners Remuneration in USD. Once Data Crawling is reached, the thesis proceeds inferring on the existence of possible correlation between financial data and Social and Blockchain data. Finally, in order to empirically evaluate the validity of the work done so far, a Multilayer Perception, a Neural Network algorithm, is rune. The forecasting performances are analyzed, computing the Mean Square Error. The work counts 5 chapters, that deeper explain the above steps and with takeaways, highlighting the fundamental concepts and results, reached in each chapter. In particular the thesis is scheduled as following: - Chapter 1: Cryptocurrency Overview - Chapter 2: Blockchian as Paradigm Shift and technology - Chapter 3: Cryptocurrency Sources and Data Crawling - Chapter 4: Heatmap and variables relationship - Chapter 5: Multilayer Perceptron application