Most　software　reliability　models　have　focused　on　merely　fault　detection　process,　and　ignored　the　time　delay　for　correcting　the　detected　faults.　Some　recent　works　considering　both　fault　detection　process　and　correction　process　are　proposed,　mainly　based　on　the　Non-homogeneous　Poisson　Process.　However,　the　analytical　models　are　usually　developed　based　on　some　specific　assumptions　on　the　fault　occurrence　time　distribution,　time　delay　distribution,　etc.　Differently,　this　paper　proposes　the　time　series　analysis　model　for　predicting　both　the　detected　number　of　faults　and　corrected　number　of　faults　during　software　testing　process.　Real　dataset　is　used　to　illustrate　the　approach.