A Study on Practical Use of Artificial Intelligence. The purpose of this research paper is to demonstrate the ease of using artificial intelligence in the light of the recent popularity of tertiary artificial intelligence. In this paper, I try to use artificial intelligence libraries that are open to the public. Additionally, I examine the prediction accuracy of machine learning, verify the recognition capability of deep learning, and discuss the necessary computer skills to use artificial intelligence. As a result of these examinations, machine learning was found to bring a rate of correct answers of more than 50%, and deep learning scored a high recognition rate. However, it is not easy for people who have never written a computer program to use artificial intelligence at this stage. Shigeki Takada JEL C87, C88 Keywords artificial intelligence, machine learning, deep learning 1. AI 1 2017 4 9 2 5 20 PONANZA 29 2 2017 5 27 AI 3, 2017a 59
71 2 37% NAUTO SHaiN CNET, 2017 1950 1960 1980 90 2000 AI,2016 AI 2 machine learning deep learning 60
,2016 Google API 2. 2.1 1 10 1 5G CATV ADSL 61
71 2 2.2 GPU GPU GPU GPU Graphics Processing Unit VRAM CPU CPU GPU CPU GPU CPU 3. machine learning 3.1 1. 2. 3. 4. 5. 6. 7. 62
3.2 scikit-learn Python Python scikit-learn Python NumPy SciPy pandas matplotlib scikit-learn scikit-learn 9 3.3 scikit-learn 2 Windows PC Windows7 8GB RAM SSD Anaconda3 Python3.6 Macbook Air MacOS 10.12.3 4GB RAM, SSD Homebrew MacOS Python2.7 3.3.1 k-db.com 2017 7 14 250 2016 CSV Excel 2016 1 4 2017 7 14 Date Open High Low Close Volume Trading Value 63
71 2 ANA KDDI ETF 6 1 3 2016 7 ETF 1 3.3.2 EFT 64
CSV 0 Excel Python 3.3.3 Random Forest Random Forest 3.3.4 Random Forest 10 ETF 2 3.3.5 1 250 249 250 65
71 2 2 ETF x x+1 x+1 +1-1 2017 7 14 20 6 16 5 19 4 18 3 21 2 20 1 23 7 250 1 23 1 22 ETF 1-1 5 5 66
3.3.6 3 51.7% 54.1% 6 16 54.2% 1 23 ANA 3 4 85.7% 1 23 5 19 6 16 83.3% 4 67
71 2 3.3.7 ETF 50% 250 4. deep learning 4.1 2 2 68
3 1940 2 4 3 4 69
71 2 4 4.2 TensorFlow TensorFlow 2015 11 Google Apache 2.0 TensorFlow deep learning 4.3 TensorFlow pixabay Python 5 66.7% 70
18.07% 5.45% 5 1 tusker, 18.07% Indian elephant, 5.45% Komodo dragon, 0.10% frilled lizard, 0.05% African elephant, 66.70% 6 89.02% 4.23% 3.09% 5 6 2 tusker, 4.23% African elephant, 3.09% Komodo dragon, 0.05% triceratops, 0.05% Indian elephant, 89.02% 7 61.34% 9.11% 7.86% 71
71 2 7 toy poodle, 61.34% Bedlington terrier, 3.30% komondor, 1.47% standard poodle, 9.11% miniature poodle, 7.86% 8 61.44% 29.44% 1.93% 8 solar dish, 1.93% notebook computer, 29.44% laptop computer, 61.44% scale, 0.34% apiary, 0.21% 11 5 8 72
5. 2 1 C++ Java Python Python 2 GPU 24 CPU 100% GPU 2 TensorFlow 73
71 2 Siri Pepper 74
Sebastian Raschka, 2016 Python impress top gear 2017 AI 14 37-53. 2016 Deep Learning Python 2015 CNET 2017 Pepper SHaiN AI https://japan.cnet.com/article/35103469/ 2017-6-29 2016 http://www.soumu.go.jp/johotsusintokei/whitepaper/ja/h28/pdf/ 28honpen.pdf 2017-5-20 2017a AI http://www.nikkei.com/article/dgxlasfg27h7b X20C17A5EA2000/ 2017-5-30 2017b http://www.nikkei.com/article/dgxlasfg20h34 Q7A520C1000000/ 2017-5-30 WEB 2017 37% https://robotstart.info/2017/07/20/sbwnauto.html 2017-7-20 URL ImageRecognition https://www.tensorflow.org/versions/master/tutorials/image recognition 2017-6-5 NVIDIA GPU http://www.nvidia.co.jp/object/what-is-gpu-computing-jp.html 2017-6-10 http://k-db.com/ 2017-7-14 Scikit-learn http://scikit-learn.org/stable/index.html 2017-6-5 https://pixabay.com/ja/ 2017-7-10 75
71 2 9 scikit-learn http://scikit-learn.org/stable/tutorial/machine learning map/ 10 11 76