The amount of information related to Hematological diseases and patients has expanded significantly. Hematological diseases should be diagnosed on time. Pathologists use conventional diagnostic tests to diagnose hematological diseases which are often low in cost and give inaccurate diagnostic results. Complete Blood Count (CBC) is used to finding the existence of diseases. Machine learning helps to elucidate acquired data and in predicting hematological diseases. Classification is used to diagnose diseases that categorize features into respective four target classes e.g., Anemia, Leukemia, Thalassemia, and healthy patients. Five machine learning algorithms were used with all features and reduced features in this investigation. The most efficient algorithm found is Random Forest with the highest accuracy at 98.59% with the lowest error rate of 0.06%. Findings show that the first indicator for blood disease is Hemoglobin.