Abstract: Stock market prediction has grown to be an interesting and intriguing research area in the field of big data analytics, predictive analytics and statistical analysis. The field of stock prediction has employed machine learning and artificial intelligence techniques to forecast the behavior of the financial market and to predict stock prices. Recently, social media has evolved to incorporate a massive amount and variety of textual data. The analysis of this information furthers the mining of public sentiment and opinions about real-time trends. In addition, the study of the inherently complex social media feeds promises new opportunities to discover empirical regularities to measure economic activity and analyze economic behavior at high frequency and in real-time. However, the noisy and short nature of social media feeds mask this information: unlike structured news content, social media content is characterized by the presence of metadata related to social media sites, (e.g., hashtags for Twitter) and the extensive usage of casual language, unstructured grammar, colloquial words, ad hoc multi-token nonstandard lexical items such as acronyms and abbreviations that need situational context to be interpreted and don’t fit with traditional technical analysis simply based on forecasting models. Under those purposes and in order to meet the trading challenge in today’s global market, technical analysis must be reconsidered. Before using any analysis model, data need to be preprocessed and regularities must be reviewed. So, the precision of the forecasting and prediction systems of the financial market and stock prices will be optimized and improved, also the accuracy of the data analysis models will be higher than state-of-art models. In this context, this study introduces the challenges of the noisy information overload from social media, gives a brief description of stock market prediction and its methodologies. Then, we discuss some of the current methods of stock prediction methodologies and emphasis the need of new improved ones which are more adapted to the context of noisy data. Finally, we present a new approach for the financial market forecasting and prediction which uses data preprocessing and normalization from noisy data in Twitter. The strong influence of the proposed data normalization model on the proposed approach’s precision and accuracy can lead to a better results than traditional ones.